{"schemaVersion":3,"updatedAt":"2026-09-17","corpusVersion":"2026-09-17.5","candidate":{"name":"Donald Shen","email":"donald7@illinois.edu","education":"Computer Engineering at the University of Illinois Urbana-Champaign","github":"https://github.com/Donaldshen27","identitySource":"Provided directly by Donald for this website.","resume":null,"graduationDate":null,"publications":[{"id":"trajectory-publication","title":"Vehicle Trajectory Prediction with Goal Estimation","year":2021,"publisher":"IEEE","venue":"2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI)","authors":["Chenxi Jin","Jiadong Shen"],"pages":"474–479","doi":"10.1109/CEI52496.2021.9574469","url":"https://doi.org/10.1109/CEI52496.2021.9574469","description":"Neural network architecture for autonomous-driving trajectory prediction, under Dr. Jiachen Li.","source":"Bibliography verified against IEEE-deposited Crossref metadata; contribution supplied by Donald."},{"id":"trade-publication","title":"Visual Analytics for the International Trade","year":2021,"publisher":"IEEE","venue":"2021 5th International Conference on Vision, Image and Signal Processing (ICVISP)","authors":["Haowen Jiang","Jiadong Shen","Qiujie Chou","Ziling Dong","Shenghui Cheng"],"pages":"296–301","doi":"10.1109/ICVISP54630.2021.00059","url":"https://doi.org/10.1109/ICVISP54630.2021.00059","description":"Mathematical modeling and data-mining visualization for international trade pattern analysis.","source":"Bibliography verified against IEEE-deposited Crossref metadata; contribution supplied by Donald."}],"experiences":[{"id":"benan-dms","organization":"Shanghai Ben’an Intelligent","role":"Research Engineer, Driver State Monitoring (DMS)","engagement":"Internship","period":"2025 – Present","location":"Remote","description":"Deployed and validated real-time on-device driver-state monitoring, connecting safety-relevant alerts, low-power edge inference, and replay-driven validation.","highlights":["Deployed and validated eight safety-relevant detector classes: eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly. Shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics alongside model development.","Designed the alert state machine with hierarchical face-loss and lens-occlusion fallbacks. Speed, ignition, and gear gating suppress false alarms in non-driving states; persisted per-camera calibration profiles support fleet-grade traceability.","Fit the model pipeline to the low-power CV181x edge SoC by tightening inference cadence and per-stage compute budgets. Validated end-to-end on real cabin video and built the lab CLI for shadow-mode hard-case capture, human review, and replay-based metric reporting."],"skills":["On-device inference","CV181x","Python","pytest","State machines","Calibration","Replay validation"],"source":"Candidate-provided internship experience and deployment update","projectIds":["driver-monitoring","dms-cv181x"]},{"id":"sjtu-research","organization":"Shanghai Jiao Tong University","role":"Autonomous Driving Research","period":"2021","description":"Implemented YOLOv3 object detection, OpenCV and SLAM, and lane-following algorithms on a ROS platform for autonomous navigation.","skills":["YOLOv3","OpenCV","SLAM","ROS","Autonomous navigation"],"source":"Candidate-provided experience"},{"id":"mckinsey-research","organization":"McKinsey & Company","role":"Part-Time Research Analyst","period":"2021","location":"Remote","description":"Produced McKinsey-standard research reports on industrial automation and inventory management systems.","skills":["Industrial automation","Inventory management","Research synthesis"],"source":"Candidate-provided experience"}],"educationExperiences":[{"id":"berkeley-summer","institution":"University of California, Berkeley","program":"Summer Session, EECS","period":"Summer 2022","description":"Summer study in electrical engineering and computer sciences.","source":"Candidate-provided education experience"}],"specialties":[{"name":"Hardware & circuits","courses":[{"code":"ECE 110","title":"Introduction to Electronics","status":"completed","group":"Hardware & circuits","skills":["Circuit measurement and modeling","Electrical circuit analysis","Sensors and motors","Electronics laboratory methods"],"fields":["embedded-vision"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-110-introduction-to-electronics-fall-2023"},{"code":"ECE 120","title":"Introduction to Computing","status":"completed","group":"Hardware & circuits","skills":["Digital logic","Combinational and sequential circuits","Finite-state machines","Computer organization and machine language"],"fields":["software-systems","embedded-vision"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-120-introduction-to-computing-spring-2024"},{"code":"ECE 210","title":"Analog Signal Processing","status":"completed","group":"Hardware & circuits","skills":["Linear circuits and systems","Convolution and stability","Laplace and Fourier transforms","Frequency response and active filters"],"fields":["embedded-vision","robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-210-analog-signal-processing-spring-2024"},{"code":"ECE 330","title":"Power Circuits & Electromechanics","status":"completed","group":"Hardware & circuits","skills":["Power and energy","Three-phase circuits","Electromagnetic forces and torques","Electric machines and energy conversion","Transducers"],"fields":["embedded-vision","robotics"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-330-power-circuits-electromechanics-fall-2025"},{"code":"ECE 342","title":"Electronic Circuits","status":"completed","group":"Hardware & circuits","skills":["Analog and digital electronic circuits","MOSFET and bipolar-transistor circuits","Amplifier analysis","Integrated-circuit design principles"],"fields":["embedded-vision"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-342-electronic-circuits-spring-2026"},{"code":"ECE 385","title":"Digital Systems Laboratory","status":"in_progress","group":"Hardware & circuits","skills":["SystemVerilog","FPGA design and verification","Timing analysis","Datapath and controller design","Hardware/software co-design"],"fields":["embedded-vision"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-385-digital-systems-laboratory-fall-2026"}],"skills":["Circuit measurement and modeling","Electrical circuit analysis","Sensors and motors","Electronics laboratory methods","Digital logic","Combinational and sequential circuits","Finite-state machines","Computer organization and machine language","Linear circuits and systems","Convolution and stability","Laplace and Fourier transforms","Frequency response and active filters","Power and energy","Three-phase circuits","Electromagnetic forces and torques","Electric machines and energy conversion","Transducers","Analog and digital electronic circuits","MOSFET and bipolar-transistor circuits","Amplifier analysis","Integrated-circuit design principles"]},{"name":"Systems & algorithms","courses":[{"code":"CS 101","title":"Introduction to Computing: Engineering & Science","status":"exam_credit","group":"Systems & algorithms","skills":["Programming and problem solving","Elementary algorithms and data structures","Scientific and engineering computation"],"fields":["software-systems"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-101-introduction-to-computing-engineering-science-credit"},{"code":"CS 124","title":"Introduction to Computer Science I","status":"transfer_credit","group":"Systems & algorithms","skills":["Fundamental programming","Computational problem solving","Introductory computing concepts"],"fields":["software-systems"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-124-introduction-to-computer-science-i-credit"},{"code":"CS 225","title":"Data Structures","status":"completed","group":"Systems & algorithms","skills":["Lists, stacks, queues and trees","Object-oriented implementation","Graph and tree search","Elementary algorithm analysis"],"fields":["software-systems"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-225-data-structures-spring-2025"},{"code":"ECE 220","title":"Computer Systems & Programming","status":"completed","group":"Systems & algorithms","skills":["LC-3 assembly and calling conventions","C programming and pointers","Dynamic memory","Recursion and elementary data structures"],"fields":["software-systems","embedded-vision"],"term":"Fall 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-220-computer-systems-programming-fall-2024"},{"code":"ECE 374","title":"Introduction to Algorithms & Models of Computation","status":"completed","group":"Systems & algorithms","skills":["Divide-and-conquer and dynamic programming","Greedy and graph algorithms","Automata and Turing machines","Reductions, undecidability and NP-completeness"],"fields":["software-systems"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"ece-374-introduction-to-algorithms-models-of-computation-spring-2026"},{"code":"ECE 391","title":"Computer Systems Engineering","status":"completed","group":"Systems & algorithms","skills":["Systems software","Input/output semantics","Synchronization and interrupts","Multitasking","Virtualization abstractions"],"fields":["software-systems","embedded-vision"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-391-computer-systems-engineering-spring-2025"},{"code":"ECE 408","title":"Applied Parallel Programming","status":"in_progress","group":"Systems & algorithms","skills":["Parallel programming models","Mapping computation to many-core hardware","Efficient parallel data structures and algorithms","Parallel application design"],"fields":["software-systems","machine-learning"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-408-applied-parallel-programming-fall-2026"},{"code":"ECE 498","title":"AI Systems & Engineering","status":"in_progress","group":"Systems & algorithms","skills":["LLM training and fine-tuning","Inference engines","AI agents and retrieval-augmented generation","PyTorch and CUDA"],"fields":["software-systems","local-llms","machine-learning"],"term":"Fall 2026","sources":["https://courses.grainger.illinois.edu/ECE498JH/fa2026/","https://courses.illinois.edu/schedule/2026/fall/ECE/498"],"enrollmentSource":"Candidate-provided enrollment update","id":"ece-498-ai-systems-engineering-fall-2026"}],"skills":["Programming and problem solving","Elementary algorithms and data structures","Scientific and engineering computation","Fundamental programming","Computational problem solving","Introductory computing concepts","Lists, stacks, queues and trees","Object-oriented implementation","Graph and tree search","Elementary algorithm analysis","LC-3 assembly and calling conventions","C programming and pointers","Dynamic memory","Recursion and elementary data structures","Divide-and-conquer and dynamic programming","Greedy and graph algorithms","Automata and Turing machines","Reductions, undecidability and NP-completeness","Systems software","Input/output semantics","Synchronization and interrupts","Multitasking","Virtualization abstractions"]},{"name":"Robotics & control","courses":[{"code":"ECE 470","title":"Introduction to Robotics","status":"completed","group":"Robotics & control","skills":["Rigid-body transformations","Forward and inverse kinematics","Motion planning and trajectories","Robotic sensing, vision and control"],"fields":["robotics"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-470-introduction-to-robotics-fall-2025"},{"code":"ECE 484","title":"Principles of Safe Autonomy","status":"completed","group":"Robotics & control","skills":["Autonomous-system perception","Modeling and motion planning","Control algorithms","Safety analysis and assumptions","Simulation and analysis tools"],"fields":["robotics","embedded-vision"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-484-principles-of-safe-autonomy-fall-2025"},{"code":"ECE 486","title":"Control Systems","status":"completed","group":"Robotics & control","skills":["Dynamic system modeling","State-space representations","Control-system analysis and design","Computational and laboratory control methods"],"fields":["robotics","embedded-vision"],"term":"Fall 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-486-control-systems-fall-2024"}],"skills":["Rigid-body transformations","Forward and inverse kinematics","Motion planning and trajectories","Robotic sensing, vision and control","Autonomous-system perception","Modeling and motion planning","Control algorithms","Safety analysis and assumptions","Simulation and analysis tools","Dynamic system modeling","State-space representations","Control-system analysis and design","Computational and laboratory control methods"]},{"name":"Machine learning","courses":[{"code":"ECE 364","title":"Programming Methods for Machine Learning","status":"completed","group":"Machine learning","skills":["Automatic differentiation","PyTorch implementation","Regression and classification","Neural-network programming","Clustering implementations"],"fields":["machine-learning","software-systems"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-364-programming-methods-for-machine-learning-spring-2025"},{"code":"ECE 449","title":"Machine Learning","status":"completed","group":"Machine learning","skills":["Supervised and generative learning","Dimensionality reduction and clustering","Neural networks","Expectation maximization","Markov decision processes and reinforcement learning"],"fields":["machine-learning"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"ece-449-machine-learning-spring-2026"},{"code":"ECE 498","title":"Deep Generative Models","status":"completed","group":"Machine learning","skills":["Variational inference and VAEs","Diffusion models","Generative adversarial networks","Normalizing flows","Mathematical foundations of generative modeling"],"fields":["machine-learning","local-llms"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/","https://courses.illinois.edu/cisapp/explorer/schedule/2025/fall/ECE/498/80771.xml"],"id":"ece-498-deep-generative-models-fall-2025"},{"code":"CS 441","title":"Applied Machine Learning","status":"completed","group":"Machine learning","skills":["Regression and classification","Clustering","Cross-validation and bootstrap","Model selection","Neural networks and applied signal problems"],"fields":["machine-learning"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-441-applied-machine-learning-spring-2026"}],"skills":["Automatic differentiation","PyTorch implementation","Regression and classification","Neural-network programming","Clustering implementations","Supervised and generative learning","Dimensionality reduction and clustering","Neural networks","Expectation maximization","Markov decision processes and reinforcement learning","Variational inference and VAEs","Diffusion models","Generative adversarial networks","Normalizing flows","Mathematical foundations of generative modeling","Clustering","Cross-validation and bootstrap","Model selection","Neural networks and applied signal problems"]},{"name":"Mathematics & optimization","courses":[{"code":"ECE 313","title":"Probability with Engineering Applications","status":"completed","group":"Mathematics & optimization","skills":["Probability theory","Reliability modeling","Statistical hypothesis testing","Decision-making under uncertainty","Parameter estimation"],"fields":["machine-learning","quant-finance"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-313-probability-with-engineering-applications-spring-2025"},{"code":"MATH 213","title":"Basic Discrete Mathematics","status":"completed","group":"Mathematics & optimization","skills":["Sets, relations and functions","Counting and recurrence relations","Graphs and trees","Algorithmic reasoning"],"fields":["software-systems"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-213-basic-discrete-mathematics-fall-2023"},{"code":"MATH 220","title":"Calculus","status":"exam_credit","group":"Mathematics & optimization","skills":["Differentiation","Integration","Fundamental theorem of calculus","Applications of single-variable calculus"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-220-calculus-credit"},{"code":"MATH 231","title":"Calculus II","status":"exam_credit","group":"Mathematics & optimization","skills":["Integration techniques","Polar coordinates","Conic sections","Infinite series"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-231-calculus-ii-credit"},{"code":"MATH 241","title":"Calculus III","status":"completed","group":"Mathematics & optimization","skills":["Partial derivatives","Multiple integrals","Vector calculus","Line and surface integrals"],"fields":["robotics","machine-learning"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-241-calculus-iii-fall-2023"},{"code":"MATH 257","title":"Linear Algebra with Computational Applications","status":"completed","group":"Mathematics & optimization","skills":["Linear systems and transformations","Eigenvalues and eigenvectors","Orthogonality and regression","Singular value decomposition","Computational linear algebra"],"fields":["machine-learning","robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-257-linear-algebra-with-computational-applications-spring-2024"},{"code":"MATH 285","title":"Introduction to Differential Equations","status":"completed","group":"Mathematics & optimization","skills":["Ordinary differential equations","Fourier series","Boundary-value problems","Introduction to partial differential equations"],"fields":["robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-285-introduction-to-differential-equations-spring-2024"},{"code":"STAT 100","title":"Statistics","status":"exam_credit","group":"Mathematics & optimization","skills":["Descriptive statistics","Elementary probability","Estimation","Hypothesis testing"],"fields":["machine-learning","quant-finance"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/stat/"],"id":"stat-100-statistics-credit"},{"code":"ECE 490","title":"Introduction to Optimization","status":"in_progress","group":"Mathematics & optimization","skills":["Unconstrained minimization","Iterative optimization methods","Linear programming","Nonlinear programming","Engineering optimization"],"fields":["machine-learning","quant-finance"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-490-introduction-to-optimization-fall-2026"}],"skills":["Probability theory","Reliability modeling","Statistical hypothesis testing","Decision-making under uncertainty","Parameter estimation","Sets, relations and functions","Counting and recurrence relations","Graphs and trees","Algorithmic reasoning","Differentiation","Integration","Fundamental theorem of calculus","Applications of single-variable calculus","Integration techniques","Polar coordinates","Conic sections","Infinite series","Partial derivatives","Multiple integrals","Vector calculus","Line and surface integrals","Linear systems and transformations","Eigenvalues and eigenvectors","Orthogonality and regression","Singular value decomposition","Computational linear algebra","Ordinary differential equations","Fourier series","Boundary-value problems","Introduction to partial differential equations","Descriptive statistics","Elementary probability","Estimation","Hypothesis testing"]},{"name":"Physical sciences","courses":[{"code":"CHEM 102","title":"General Chemistry I","status":"exam_credit","group":"Physical sciences","skills":["Atomic structure and bonding","States of matter","Stoichiometry","Chemical equilibrium"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/chem/"],"id":"chem-102-general-chemistry-i-credit"},{"code":"CHEM 104","title":"General Chemistry II","status":"exam_credit","group":"Physical sciences","skills":["Chemical energetics","Kinetics and equilibrium","Electrochemistry","Chemistry of materials"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/chem/"],"id":"chem-104-general-chemistry-ii-credit"},{"code":"PHYS 102","title":"College Physics: Electricity, Magnetism & Modern Physics","status":"exam_credit","group":"Physical sciences","skills":["Electric and magnetic fields","Basic circuits","Geometrical optics","Introductory modern physics"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-102-college-physics-electricity-magnetism-modern-physics-credit"},{"code":"PHYS 211","title":"University Physics: Mechanics","status":"completed","group":"Physical sciences","skills":["Newtonian mechanics","Work and energy","Rotational dynamics","Oscillations and waves"],"fields":["robotics"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-211-university-physics-mechanics-fall-2023"},{"code":"PHYS 212","title":"University Physics: Electricity & Magnetism","status":"exam_credit","group":"Physical sciences","skills":["Electric fields and Gauss law","Capacitance and circuits","Magnetic fields and induction","Electromagnetic waves and optics"],"fields":["embedded-vision"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-212-university-physics-electricity-magnetism-credit"},{"code":"PHYS 213","title":"University Physics: Thermal Physics","status":"completed","group":"Physical sciences","skills":["Thermodynamics","Kinetic theory","Entropy and statistical mechanics","Free energy and Boltzmann factors"],"fields":[],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-213-university-physics-thermal-physics-spring-2024"},{"code":"PHYS 214","title":"University Physics: Quantum Physics","status":"completed","group":"Physical sciences","skills":["Interference and diffraction","Photons and matter waves","Atomic models","Uncertainty and wave mechanics"],"fields":["embedded-vision"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-214-university-physics-quantum-physics-spring-2024"}],"skills":["Atomic structure and bonding","States of matter","Stoichiometry","Chemical equilibrium","Chemical energetics","Kinetics and equilibrium","Electrochemistry","Chemistry of materials","Electric and magnetic fields","Basic circuits","Geometrical optics","Introductory modern physics","Newtonian mechanics","Work and energy","Rotational dynamics","Oscillations and waves","Electric fields and Gauss law","Capacitance and circuits","Magnetic fields and induction","Electromagnetic waves and optics","Thermodynamics","Kinetic theory","Entropy and statistical mechanics","Free energy and Boltzmann factors","Interference and diffraction","Photons and matter waves","Atomic models","Uncertainty and wave mechanics"]},{"name":"Economics","courses":[{"code":"ECON 103","title":"Macroeconomic Principles","status":"exam_credit","group":"Economics","skills":["Aggregate income, employment and output","Money and price levels","Monetary and fiscal policy","Inflation, unemployment and economic growth","International economics"],"term":null,"fields":["quant-finance"],"sources":["https://catalog.illinois.edu/courses-of-instruction/econ/"],"id":"econ-103-macroeconomic-principles-credit"},{"code":"IE 421","title":"High Frequency Trading Technology","status":"in_progress","group":"Economics","skills":["Automated trading mechanics","Real-time price formation","Market-data processing","High-speed trade execution"],"fields":["quant-finance","software-systems"],"term":"Fall 2026","sources":["https://courses.illinois.edu/schedule/2026/fall/IE/421","https://catalog.illinois.edu/courses-of-instruction/ie/"],"enrollmentSource":"Candidate-provided enrollment update","id":"ie-421-high-frequency-trading-technology-fall-2026"}],"skills":["Aggregate income, employment and output","Money and price levels","Monetary and fiscal policy","Inflation, unemployment and economic growth","International economics"]}],"coursework":[{"code":"ECE 110","title":"Introduction to Electronics","status":"completed","group":"Hardware & circuits","skills":["Circuit measurement and modeling","Electrical circuit analysis","Sensors and motors","Electronics laboratory methods"],"fields":["embedded-vision"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-110-introduction-to-electronics-fall-2023"},{"code":"ECE 120","title":"Introduction to Computing","status":"completed","group":"Hardware & circuits","skills":["Digital logic","Combinational and sequential circuits","Finite-state machines","Computer organization and machine language"],"fields":["software-systems","embedded-vision"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-120-introduction-to-computing-spring-2024"},{"code":"ECE 210","title":"Analog Signal Processing","status":"completed","group":"Hardware & circuits","skills":["Linear circuits and systems","Convolution and stability","Laplace and Fourier transforms","Frequency response and active filters"],"fields":["embedded-vision","robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-210-analog-signal-processing-spring-2024"},{"code":"ECE 330","title":"Power Circuits & Electromechanics","status":"completed","group":"Hardware & circuits","skills":["Power and energy","Three-phase circuits","Electromagnetic forces and torques","Electric machines and energy conversion","Transducers"],"fields":["embedded-vision","robotics"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-330-power-circuits-electromechanics-fall-2025"},{"code":"ECE 342","title":"Electronic Circuits","status":"completed","group":"Hardware & circuits","skills":["Analog and digital electronic circuits","MOSFET and bipolar-transistor circuits","Amplifier analysis","Integrated-circuit design principles"],"fields":["embedded-vision"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-342-electronic-circuits-spring-2026"},{"code":"ECE 385","title":"Digital Systems Laboratory","status":"in_progress","group":"Hardware & circuits","skills":["SystemVerilog","FPGA design and verification","Timing analysis","Datapath and controller design","Hardware/software co-design"],"fields":["embedded-vision"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-385-digital-systems-laboratory-fall-2026"},{"code":"CS 101","title":"Introduction to Computing: Engineering & Science","status":"exam_credit","group":"Systems & algorithms","skills":["Programming and problem solving","Elementary algorithms and data structures","Scientific and engineering computation"],"fields":["software-systems"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-101-introduction-to-computing-engineering-science-credit"},{"code":"CS 124","title":"Introduction to Computer Science I","status":"transfer_credit","group":"Systems & algorithms","skills":["Fundamental programming","Computational problem solving","Introductory computing concepts"],"fields":["software-systems"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-124-introduction-to-computer-science-i-credit"},{"code":"CS 225","title":"Data Structures","status":"completed","group":"Systems & algorithms","skills":["Lists, stacks, queues and trees","Object-oriented implementation","Graph and tree search","Elementary algorithm analysis"],"fields":["software-systems"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-225-data-structures-spring-2025"},{"code":"ECE 220","title":"Computer Systems & Programming","status":"completed","group":"Systems & algorithms","skills":["LC-3 assembly and calling conventions","C programming and pointers","Dynamic memory","Recursion and elementary data structures"],"fields":["software-systems","embedded-vision"],"term":"Fall 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-220-computer-systems-programming-fall-2024"},{"code":"ECE 374","title":"Introduction to Algorithms & Models of Computation","status":"completed","group":"Systems & algorithms","skills":["Divide-and-conquer and dynamic programming","Greedy and graph algorithms","Automata and Turing machines","Reductions, undecidability and NP-completeness"],"fields":["software-systems"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"ece-374-introduction-to-algorithms-models-of-computation-spring-2026"},{"code":"ECE 391","title":"Computer Systems Engineering","status":"completed","group":"Systems & algorithms","skills":["Systems software","Input/output semantics","Synchronization and interrupts","Multitasking","Virtualization abstractions"],"fields":["software-systems","embedded-vision"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-391-computer-systems-engineering-spring-2025"},{"code":"ECE 408","title":"Applied Parallel Programming","status":"in_progress","group":"Systems & algorithms","skills":["Parallel programming models","Mapping computation to many-core hardware","Efficient parallel data structures and algorithms","Parallel application design"],"fields":["software-systems","machine-learning"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-408-applied-parallel-programming-fall-2026"},{"code":"ECE 470","title":"Introduction to Robotics","status":"completed","group":"Robotics & control","skills":["Rigid-body transformations","Forward and inverse kinematics","Motion planning and trajectories","Robotic sensing, vision and control"],"fields":["robotics"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-470-introduction-to-robotics-fall-2025"},{"code":"ECE 484","title":"Principles of Safe Autonomy","status":"completed","group":"Robotics & control","skills":["Autonomous-system perception","Modeling and motion planning","Control algorithms","Safety analysis and assumptions","Simulation and analysis tools"],"fields":["robotics","embedded-vision"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-484-principles-of-safe-autonomy-fall-2025"},{"code":"ECE 486","title":"Control Systems","status":"completed","group":"Robotics & control","skills":["Dynamic system modeling","State-space representations","Control-system analysis and design","Computational and laboratory control methods"],"fields":["robotics","embedded-vision"],"term":"Fall 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-486-control-systems-fall-2024"},{"code":"ECE 364","title":"Programming Methods for Machine Learning","status":"completed","group":"Machine learning","skills":["Automatic differentiation","PyTorch implementation","Regression and classification","Neural-network programming","Clustering implementations"],"fields":["machine-learning","software-systems"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-364-programming-methods-for-machine-learning-spring-2025"},{"code":"ECE 449","title":"Machine Learning","status":"completed","group":"Machine learning","skills":["Supervised and generative learning","Dimensionality reduction and clustering","Neural networks","Expectation maximization","Markov decision processes and reinforcement learning"],"fields":["machine-learning"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"ece-449-machine-learning-spring-2026"},{"code":"ECE 498","title":"Deep Generative Models","status":"completed","group":"Machine learning","skills":["Variational inference and VAEs","Diffusion models","Generative adversarial networks","Normalizing flows","Mathematical foundations of generative modeling"],"fields":["machine-learning","local-llms"],"term":"Fall 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/","https://courses.illinois.edu/cisapp/explorer/schedule/2025/fall/ECE/498/80771.xml"],"id":"ece-498-deep-generative-models-fall-2025"},{"code":"ECE 313","title":"Probability with Engineering Applications","status":"completed","group":"Mathematics & optimization","skills":["Probability theory","Reliability modeling","Statistical hypothesis testing","Decision-making under uncertainty","Parameter estimation"],"fields":["machine-learning","quant-finance"],"term":"Spring 2025","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-313-probability-with-engineering-applications-spring-2025"},{"code":"MATH 213","title":"Basic Discrete Mathematics","status":"completed","group":"Mathematics & optimization","skills":["Sets, relations and functions","Counting and recurrence relations","Graphs and trees","Algorithmic reasoning"],"fields":["software-systems"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-213-basic-discrete-mathematics-fall-2023"},{"code":"MATH 220","title":"Calculus","status":"exam_credit","group":"Mathematics & optimization","skills":["Differentiation","Integration","Fundamental theorem of calculus","Applications of single-variable calculus"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-220-calculus-credit"},{"code":"MATH 231","title":"Calculus II","status":"exam_credit","group":"Mathematics & optimization","skills":["Integration techniques","Polar coordinates","Conic sections","Infinite series"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-231-calculus-ii-credit"},{"code":"MATH 241","title":"Calculus III","status":"completed","group":"Mathematics & optimization","skills":["Partial derivatives","Multiple integrals","Vector calculus","Line and surface integrals"],"fields":["robotics","machine-learning"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-241-calculus-iii-fall-2023"},{"code":"MATH 257","title":"Linear Algebra with Computational Applications","status":"completed","group":"Mathematics & optimization","skills":["Linear systems and transformations","Eigenvalues and eigenvectors","Orthogonality and regression","Singular value decomposition","Computational linear algebra"],"fields":["machine-learning","robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-257-linear-algebra-with-computational-applications-spring-2024"},{"code":"MATH 285","title":"Introduction to Differential Equations","status":"completed","group":"Mathematics & optimization","skills":["Ordinary differential equations","Fourier series","Boundary-value problems","Introduction to partial differential equations"],"fields":["robotics"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/math/"],"id":"math-285-introduction-to-differential-equations-spring-2024"},{"code":"STAT 100","title":"Statistics","status":"exam_credit","group":"Mathematics & optimization","skills":["Descriptive statistics","Elementary probability","Estimation","Hypothesis testing"],"fields":["machine-learning","quant-finance"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/stat/"],"id":"stat-100-statistics-credit"},{"code":"ECE 490","title":"Introduction to Optimization","status":"in_progress","group":"Mathematics & optimization","skills":["Unconstrained minimization","Iterative optimization methods","Linear programming","Nonlinear programming","Engineering optimization"],"fields":["machine-learning","quant-finance"],"term":"Fall 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/ece/"],"id":"ece-490-introduction-to-optimization-fall-2026"},{"code":"CHEM 102","title":"General Chemistry I","status":"exam_credit","group":"Physical sciences","skills":["Atomic structure and bonding","States of matter","Stoichiometry","Chemical equilibrium"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/chem/"],"id":"chem-102-general-chemistry-i-credit"},{"code":"CHEM 104","title":"General Chemistry II","status":"exam_credit","group":"Physical sciences","skills":["Chemical energetics","Kinetics and equilibrium","Electrochemistry","Chemistry of materials"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/chem/"],"id":"chem-104-general-chemistry-ii-credit"},{"code":"CS 441","title":"Applied Machine Learning","status":"completed","group":"Machine learning","skills":["Regression and classification","Clustering","Cross-validation and bootstrap","Model selection","Neural networks and applied signal problems"],"fields":["machine-learning"],"term":"Spring 2026","sources":["https://catalog.illinois.edu/courses-of-instruction/cs/"],"id":"cs-441-applied-machine-learning-spring-2026"},{"code":"PHYS 102","title":"College Physics: Electricity, Magnetism & Modern Physics","status":"exam_credit","group":"Physical sciences","skills":["Electric and magnetic fields","Basic circuits","Geometrical optics","Introductory modern physics"],"fields":[],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-102-college-physics-electricity-magnetism-modern-physics-credit"},{"code":"PHYS 211","title":"University Physics: Mechanics","status":"completed","group":"Physical sciences","skills":["Newtonian mechanics","Work and energy","Rotational dynamics","Oscillations and waves"],"fields":["robotics"],"term":"Fall 2023","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-211-university-physics-mechanics-fall-2023"},{"code":"PHYS 212","title":"University Physics: Electricity & Magnetism","status":"exam_credit","group":"Physical sciences","skills":["Electric fields and Gauss law","Capacitance and circuits","Magnetic fields and induction","Electromagnetic waves and optics"],"fields":["embedded-vision"],"term":null,"sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-212-university-physics-electricity-magnetism-credit"},{"code":"PHYS 213","title":"University Physics: Thermal Physics","status":"completed","group":"Physical sciences","skills":["Thermodynamics","Kinetic theory","Entropy and statistical mechanics","Free energy and Boltzmann factors"],"fields":[],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-213-university-physics-thermal-physics-spring-2024"},{"code":"PHYS 214","title":"University Physics: Quantum Physics","status":"completed","group":"Physical sciences","skills":["Interference and diffraction","Photons and matter waves","Atomic models","Uncertainty and wave mechanics"],"fields":["embedded-vision"],"term":"Spring 2024","sources":["https://catalog.illinois.edu/courses-of-instruction/phys/"],"id":"phys-214-university-physics-quantum-physics-spring-2024"},{"code":"ECON 103","title":"Macroeconomic Principles","status":"exam_credit","group":"Economics","skills":["Aggregate income, employment and output","Money and price levels","Monetary and fiscal policy","Inflation, unemployment and economic growth","International economics"],"term":null,"fields":["quant-finance"],"sources":["https://catalog.illinois.edu/courses-of-instruction/econ/"],"id":"econ-103-macroeconomic-principles-credit"},{"code":"ECE 498","title":"AI Systems & Engineering","status":"in_progress","group":"Systems & algorithms","skills":["LLM training and fine-tuning","Inference engines","AI agents and retrieval-augmented generation","PyTorch and CUDA"],"fields":["software-systems","local-llms","machine-learning"],"term":"Fall 2026","sources":["https://courses.grainger.illinois.edu/ECE498JH/fa2026/","https://courses.illinois.edu/schedule/2026/fall/ECE/498"],"enrollmentSource":"Candidate-provided enrollment update","id":"ece-498-ai-systems-engineering-fall-2026"},{"code":"IE 421","title":"High Frequency Trading Technology","status":"in_progress","group":"Economics","skills":["Automated trading mechanics","Real-time price formation","Market-data processing","High-speed trade execution"],"fields":["quant-finance","software-systems"],"term":"Fall 2026","sources":["https://courses.illinois.edu/schedule/2026/fall/IE/421","https://catalog.illinois.edu/courses-of-instruction/ie/"],"enrollmentSource":"Candidate-provided enrollment update","id":"ie-421-high-frequency-trading-technology-fall-2026"}],"voice":[]},"fields":[{"id":"robotics","name":"Robotics & Autonomous Driving","shortName":"Robotics & autonomy","description":"Localization, navigation, vehicle control, and the physical consequences of a decision.","question":"How does Donald connect perception to action?"},{"id":"embedded-vision","name":"Embedded Systems & Computer Vision","shortName":"Embedded systems & vision","description":"Driver monitoring, constrained-device inference, and reconstructing the world from images.","question":"What has Donald built in driver monitoring and embedded vision?"},{"id":"machine-learning","name":"Machine Learning & Model Research","shortName":"Machine learning","description":"Forecasting, representation learning, reinforcement learning, and experiments inside the model.","question":"Which machine learning projects show the most technical depth?"},{"id":"local-llms","name":"Local LLM Inference & Agents","shortName":"Local LLMs & agents","description":"Running models locally, adapting them, and giving agent workflows an accountable structure.","question":"What has Donald worked on in local LLM inference?"},{"id":"software-systems","name":"Software Systems & Automation","shortName":"Systems & automation","description":"Industrial SOP verification, concurrency runtimes, and tools that turn a workflow into a system.","question":"What is the SOP verification system, and how does it work?"},{"id":"quant-finance","name":"Quantitative Finance & Markets","shortName":"Quantitative finance","description":"Execution, statistical arbitrage, market mechanisms, and research that makes its assumptions visible.","question":"How do Donald’s market projects differ from one another?"},{"id":"creative-tools","name":"Interactive Tools, Learning & Writing","shortName":"Tools, learning & writing","description":"A soccer pitch, a piano roll, a knowledge graph, and a study of technology beyond the code.","question":"Show me Donald’s work outside trading and AI research."}],"rankingMethod":"Within each field, projects are ordered by engineering depth, scope of Donald’s contribution, strength of the available evidence, and how clearly they show a distinct capability. This is an editorial judgment, not a performance score. Related repositories are grouped into project families.","projects":[{"id":"autonomous-driving-lab","date":"Project record · Sep 2026","url":"/projects/autonomous-driving-lab","visual":"schema","why":"A ROS2/Gazebo course project connecting lane segmentation, curvature-aware speed control, and LiDAR/GPS localization.","question":"","hard":"Lane masks transformed into bird-eye geometry and fitted lane curves; waypoint lookahead and turn anticipation for speed selection; particle likelihoods with finite-weight safeguards, systematic resampling and GPS reseeding under low effective sample size; eight-direction LiDAR support; offline generated motion/noise trials measuring position RMSE, heading RMSE, convergence and ESS.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Shared course/team workspace; do not attribute the simulator, model backbone or all source to Donald. Configured rates and speeds are not measured physical performance. Offline tuning uses synthetic sequences and cannot substitute for physical vehicle evaluation. MP0 contains a preliminary decision-rule study with remaining scaffolding; it is not a separately verified autonomous system.","attribution":"Donald implemented lane-segmentation stages, curvature-based speed control, particle-filter updates, eight-direction LiDAR support, and an offline localization parameter sweep within shared course infrastructure.","contribution":"Donald implemented lane-segmentation stages, curvature-based speed control, particle-filter updates, eight-direction LiDAR support, and an offline localization parameter sweep within shared course infrastructure.","result":"An integrated set of perception, control, and localization modules, with training checkpoints and simulation experiments that compare position error, heading error, convergence, and effective sample size.","name":"Autonomous driving lab","title":"Autonomous driving lab","description":"A ROS2/Gazebo course project connecting lane segmentation, curvature-aware speed control, and LiDAR/GPS localization.","category":"Robotics & Autonomous Driving","technologies":["Python","ROS2","Gazebo","PyTorch","SimpleENet","Particle filters","LiDAR","Vehicle control"],"roles":["Robotics / Autonomy","Research"],"evidence":["autonomous-driving-lab-catalog-1","autonomous-driving-lab-catalog-2","autonomous-driving-lab-catalog-3","autonomous-driving-lab-catalog-4","autonomous-driving-lab-catalog-5","autonomous-driving-lab-catalog-6"],"field":"robotics","rank":1,"rankReason":"The broadest autonomy work: perception, control, localization, and explicit evaluation in one course sequence.","folders":["uiuc/ece484"],"kind":"study","fullStory":false,"limitations":["Shared course/team workspace; do not attribute the simulator, model backbone or all source to Donald.","Configured rates and speeds are not measured physical performance.","Offline tuning uses synthetic sequences and cannot substitute for physical vehicle evaluation.","MP0 contains a preliminary decision-rule study with remaining scaffolding; it is not a separately verified autonomous system."]},{"id":"swarm-navigation","date":"December 2025","url":"/projects/swarm-navigation","visual":"swarm","why":"Navigate five robots around static obstacles, one another, and a bounded environment across every goal-assignment permutation in a fixed study.","question":"How do attraction, repulsion, and circulation cooperate without trapping the robots?","hard":"Too much goal attraction can overpower repulsion. Local minima and orbiting can prevent completion.","assumptions":"Robots and obstacles follow the course simulator’s geometry and dynamics. This is a fixed benchmark, not a general proof.","data":"The saved results table contains 120 assignments; all 120 report successful completion. Mean completion was independently recomputed as 8,434.06 simulation steps.","failure":"The report documents orbiting, local minima, and collisions during parameter exploration.","changed":"The controller combines nonlinear attraction, obstacle circulation, robot repulsion, and conditional boundary repulsion.","lessons":"Evaluate the assignments where robots interfere, not only visually clean trajectories.","unresolved":"The reactive controller does not anticipate future conflicts. Deadlock detection and velocity-aware avoidance are proposed future work.","attribution":"Donald-attributed ECE470 course project, using supplied simulator infrastructure.","name":"Multi-robot navigation","title":"Multi-robot navigation","description":"A potential-field controller evaluated over every goal assignment for five robots in a fixed obstacle-filled circular workspace.","category":"Robotics & Autonomous Driving","technologies":["Python","Control","Simulation","Potential fields","NumPy"],"roles":["Research","Software Engineering","AI / ML","Robotics / Autonomy"],"contribution":"Donald developed and tuned the potential-field controller, parameter sweeps, and trajectory visualizations for an ECE470 project using a supplied simulator.","result":"All 120 goal assignments completed in the saved five-robot simulation suite, averaging 8,434 simulation steps. The result applies to that fixed simulator and obstacle configuration.","evidence":["swarm-code","swarm-results","swarm-failure","swarm-navigation-catalog-1","swarm-navigation-catalog-2","swarm-navigation-catalog-3","swarm-navigation-catalog-4"],"field":"robotics","rank":3,"rankReason":"Best finite quantitative outcome evidence in the course set, though a smaller and more constrained project than the driving systems.","folders":["uiuc/ECE_470_FA25_Project-Code"],"kind":"study","fullStory":true,"limitations":["Fixed course simulation suite, not a general convergence guarantee or physical-robot result.","Prose report min/max differ from the saved CSV; use only checked 120/120 and mean.","Excessive attraction caused collisions; nonlinear attraction addressed orbiting/local minima.","The controller remains reactive and does not anticipate future conflicts."]},{"id":"f1tenth","date":"November 2025","url":"/projects/f1tenth","visual":"control","why":"Connect camera-derived lane geometry to a control path while handling missing or stale perception.","question":"What should a controller do when its input no longer describes a trustworthy path?","hard":"Pixel geometry must become robot-frame geometry. Confidence and age matter when choosing a path.","assumptions":"Camera calibration, timestamps, and frame transforms must remain consistent. Configured speeds are targets, not measurements.","data":"Camera lane features and ROS2 path messages. Physical track results were not verified.","failure":"Missing paths and lookahead failures need explicit stop behavior rather than continued motion on stale information.","changed":"Safety fixes stop camera-only control when no valid path is available.","lessons":"An unavailable perception result needs an intentional control response.","unresolved":"Physical deployment performance, calibration robustness, and latency require validation.","attribution":"Donald’s verified additions to a shared, upstream F1Tenth course workspace.","name":"F1Tenth lane following","title":"F1Tenth lane following","description":"A camera-to-control pipeline for a F1Tenth ROS2 platform, with confidence-aware path selection and explicit stop behavior when vision has no usable path.","category":"Robotics & Autonomous Driving","technologies":["Python","ROS2","Computer vision","Pure pursuit","OpenCV","RealSense","Homography"],"roles":["Software Engineering","AI / ML","Research","Robotics / Autonomy"],"contribution":"Donald added camera-based lane extraction, configurable image-to-robot coordinate conversion, confidence-aware path selection, and missing-path and lookahead stop behavior to an existing F1Tenth course workspace.","result":"A perception-to-control implementation that selects paths by confidence and freshness, then stops camera-only control when no usable path remains.","evidence":["control-stop","f1tenth-catalog-1","f1tenth-catalog-2","f1tenth-catalog-3","f1tenth-catalog-4"],"field":"robotics","rank":4,"rankReason":"Clear candidate contributions and physical-system reasoning, with narrower scope and unverified hardware outcomes.","folders":["uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws","uiuc/ece484_f1_tenth_cr7_ws/f1tenth_ws"],"kind":"adaptation","fullStory":true,"limitations":["Existing UIUC Robotics/F1Tenth foundation with multiple contributors.","Nominal camera rates, control rates and speeds are configurations rather than measurements.","No claim of a completed physical track trial or real-world reliability.","Physical track performance is still unverified."]},{"id":"ur3-manipulation","date":"Project record · Sep 2026","url":"/projects/ur3-manipulation","visual":"schema","why":"Course exercises for a UR3 arm: recursive Tower of Hanoi planning, suction and motion coordination, plus an unfinished vision-based block-manipulation extension.","question":"","hard":"ROS command/feedback loops; recursive task planning; arm and vacuum control; calibrated coordinate transforms; morphological cleanup and blob filtering for colored blocks.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"No physical execution success or author-by-line provenance verified. Lab5 Get_MS returns identity/zero scaffolding and inverse kinematics is unimplemented; do not present it as completed vision pick-and-place. Course drivers and starter code are upstream.","attribution":"Student-code implementations add recursive three-block planning, state-dependent pickup and placement heights, suction feedback, colored-block detection, and image-to-world transforms to supplied course scaffolding.","contribution":"Student-code implementations add recursive three-block planning, state-dependent pickup and placement heights, suction feedback, colored-block detection, and image-to-world transforms to supplied course scaffolding.","result":"The Hanoi implementation combines a recursive planner with calibrated arm positions. The vision extension includes blob extraction but still needs its kinematics and calibration/goal placeholders completed.","name":"UR3 manipulation & vision labs","title":"UR3 manipulation & vision labs","description":"Course exercises for a UR3 arm: recursive Tower of Hanoi planning, suction and motion coordination, plus an unfinished vision-based block-manipulation extension.","category":"Robotics & Autonomous Driving","technologies":["Python","ROS","UR3","OpenCV","Kinematics"],"roles":["Robotics / Autonomy","Research"],"evidence":["ur3-manipulation-catalog-1","ur3-manipulation-catalog-2","ur3-manipulation-catalog-3","ur3-manipulation-catalog-4"],"field":"robotics","rank":5,"rankReason":"Meaningful learning artifact with implementation, but unfinished subcomponents and weaker attribution require lower emphasis.","folders":["uiuc/ece470"],"kind":"study","fullStory":false,"limitations":["No physical execution success or author-by-line provenance verified.","Lab5 Get_MS returns identity/zero scaffolding and inverse kinematics is unimplemented; do not present it as completed vision pick-and-place.","Course drivers and starter code are upstream."]},{"id":"driver-monitoring","date":"2025 – Present","url":"/projects/driver-monitoring","location":"Remote","experienceId":"benan-dms","visual":"schema","why":"On-device driver-state monitoring developed during Donald’s research-engineering internship at Shanghai Ben’an Intelligent: eight detector classes, vehicle-aware alerts, and replay-driven validation.","question":"","hard":"Eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly detection; hierarchical face-loss and lens-occlusion fallbacks; speed, ignition, and gear gating; per-camera calibration; shadow-mode hard-case capture, human review, and replay-based metrics.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Employment, on-device deployment, and cabin-video validation reflect Donald’s latest internship record. The 616-test suite is a project result, not a new website-run test. The supplied record does not claim safety certification, verified CAN integration, or quantified fleet-wide sensitivity/specificity.","attribution":"Deployed and validated real-time inference for eight safety-relevant detector classes. Designed the hierarchical alert state machine and vehicle-context gating, persisted per-camera calibration profiles, and built the diagnostic lab CLI.","contribution":"Deployed and validated real-time inference for eight safety-relevant detector classes. Designed the hierarchical alert state machine and vehicle-context gating, persisted per-camera calibration profiles, and built the diagnostic lab CLI.","result":"Validated on real cabin video and shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics in lockstep with model development.","name":"Driver monitoring & diagnostic lab","title":"Driver monitoring & diagnostic lab","description":"On-device driver-state monitoring developed during Donald’s research-engineering internship at Shanghai Ben’an Intelligent: eight detector classes, vehicle-aware alerts, and replay-driven validation.","category":"Embedded Systems & Computer Vision","technologies":["Python","MediaPipe","OpenCV","Temporal state machines","Calibration","Replay tooling"],"roles":["Embedded / Vision","Robotics / Autonomy","AI / ML"],"evidence":["benan-dms","driver-monitoring-catalog-1","driver-monitoring-catalog-2","driver-monitoring-catalog-3","driver-monitoring-catalog-4","driver-monitoring-catalog-5"],"field":"embedded-vision","rank":1,"rankReason":"An end-to-end internship engineering story: on-device deployment, context-aware alerts, calibration traceability, and a 616-test validation suite.","folders":["dms"],"kind":"built","fullStory":false,"limitations":["Employment, on-device deployment, and cabin-video validation reflect Donald’s latest internship record. The 616-test suite is a project result, not a new website-run test.","The supplied record does not claim safety certification, verified CAN integration, or quantified fleet-wide sensitivity/specificity."]},{"id":"dms-cv181x","date":"2025 – Present","url":"/projects/dms-cv181x","location":"Remote","experienceId":"benan-dms","visual":"schema","why":"A C inference pipeline for constrained hardware: sparse face detection, tracked regions, lightweight landmarks, and per-stage timing.","question":"","hard":"Sparse SCRFD detection with per-frame landmark tracking and ROI updates; legacy 478-point and 68-point backends; callbacks and display work moved off the per-frame critical path; timestamp-window PERCLOS; per-stage latency and tracking instrumentation; ONNX to CV181x BF16 conversion.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"The latest candidate-provided internship record updates the earlier repository-only hardware-validation status. It describes on-device deployment and end-to-end cabin-video validation; no numeric latency or accuracy benchmark is supplied. 20 ms remains a target, not a claimed achieved measurement. SCRFD and PFLD are upstream models; original model training is not claimed. Company-specific source is not mirrored on this website. The code computes sample ratios inside a timestamp window; do not describe it as exact time-integrated PERCLOS.","attribution":"Fit the DMS model pipeline to the low-power CV181x edge SoC during the Shanghai Ben’an Intelligent internship. Tightened inference cadence and per-stage compute budgets while adapting a supplied C service, landmark backends, and model-conversion packaging.","contribution":"Fit the DMS model pipeline to the low-power CV181x edge SoC during the Shanghai Ben’an Intelligent internship. Tightened inference cadence and per-stage compute budgets while adapting a supplied C service, landmark backends, and model-conversion packaging.","result":"Validated the end-to-end pipeline on real cabin video, with inference-cadence checks and per-stage compute budgeting. The internship record connects this edge pipeline to shadow-mode hard-case capture, review, and replay-based reporting.","name":"DMS on CV181x","title":"DMS on CV181x","description":"A C inference pipeline for constrained hardware: sparse face detection, tracked regions, lightweight landmarks, and per-stage timing.","category":"Embedded Systems & Computer Vision","technologies":["C","CV181x","SCRFD","PFLD","TPU-MLIR","ONNX","Embedded inference"],"roles":["Embedded / Vision","Robotics / Autonomy","AI / ML"],"evidence":["benan-dms","dms-cv181x-catalog-1","dms-cv181x-catalog-2","dms-cv181x-catalog-3","dms-cv181x-catalog-4"],"field":"embedded-vision","rank":2,"rankReason":"Hardware-aware internship work connecting a low-power edge pipeline to real cabin-video validation and replay-driven engineering.","folders":["dms_cv181x_fastpath/dms_work"],"kind":"adaptation","fullStory":false,"limitations":["The latest candidate-provided internship record updates the earlier repository-only hardware-validation status. It describes on-device deployment and end-to-end cabin-video validation; no numeric latency or accuracy benchmark is supplied.","20 ms remains a target, not a claimed achieved measurement.","SCRFD and PFLD are upstream models; original model training is not claimed.","Company-specific source is not mirrored on this website.","The code computes sample ratios inside a timestamp window; do not describe it as exact time-integrated PERCLOS."]},{"id":"vggt-reconstruction","date":"Project record · Sep 2026","url":"/projects/vggt-reconstruction","visual":"schema","why":"A local adaptation of VGGT that filters sky regions from reconstructed point clouds and adds reconstruction inspection tools.","question":"","hard":"ONNX sky segmentation, VGGT depth and camera prediction, depth unprojection, point filtering, and interactive inspection of the reconstructed geometry.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"The original VGGT architecture and model weights are upstream. Local script presence does not establish a new model or a quantitative reconstruction result. Reconstruction quality and physical-scale accuracy remain unbenchmarked.","attribution":"Local additions wrap upstream VGGT with ONNX sky-mask preprocessing, confidence-based point filtering, reconstruction export, and geometry inspection tools.","contribution":"Local additions wrap upstream VGGT with ONNX sky-mask preprocessing, confidence-based point filtering, reconstruction export, and geometry inspection tools.","result":"Reconstruction and visualization scripts produce inspectable point-cloud artifacts with sky filtering.","name":"VGGT scene reconstruction","title":"VGGT scene reconstruction","description":"A local adaptation of VGGT that filters sky regions from reconstructed point clouds and adds reconstruction inspection tools.","category":"Embedded Systems & Computer Vision","technologies":["PyTorch","VGGT","ONNX Runtime","3D geometry","Viser"],"roles":["Embedded / Vision","Robotics / Autonomy","AI / ML"],"evidence":["vggt-reconstruction-catalog-1","vggt-reconstruction-catalog-2","vggt-reconstruction-catalog-3"],"field":"embedded-vision","rank":3,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["vggt"],"kind":"adaptation","fullStory":false,"limitations":["The original VGGT architecture and model weights are upstream. Local script presence does not establish a new model or a quantitative reconstruction result.","Reconstruction quality and physical-scale accuracy remain unbenchmarked."]},{"id":"sam3-experiments","date":"Project record · Sep 2026","url":"/projects/sam3-experiments","visual":"schema","why":"Prompt-based image-segmentation experiments in Meta’s SAM 3 notebook, with local environment and prompt changes.","question":"","hard":"Text and geometric prompts, confidence thresholds, image-mask inspection, and CUDA mixed-precision inference in the supplied notebook.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"This is an upstream-model experiment, not original SAM 3 development. Changes are currently uncommitted and no independent performance evaluation was inspected.","attribution":"A small hands-on study modifying example prompts and Python compatibility settings in Meta’s supplied SAM 3 notebook. The model and predictor are upstream implementations.","contribution":"A small hands-on study modifying example prompts and Python compatibility settings in Meta’s supplied SAM 3 notebook. The model and predictor are upstream implementations.","result":"A locally adapted notebook for inspecting prompt-based segmentation, without a new training or benchmark result.","name":"SAM 3 segmentation experiments","title":"SAM 3 segmentation experiments","description":"Prompt-based image-segmentation experiments in Meta’s SAM 3 notebook, with local environment and prompt changes.","category":"Embedded Systems & Computer Vision","technologies":["SAM 3","PyTorch","Image segmentation","CUDA","Jupyter"],"roles":["Embedded / Vision","Robotics / Autonomy","AI / ML"],"evidence":["sam3-experiments-catalog-1","sam3-experiments-catalog-2"],"field":"embedded-vision","rank":4,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["sam3"],"kind":"study","fullStory":false,"limitations":["This is an upstream-model experiment, not original SAM 3 development. Changes are currently uncommitted and no independent performance evaluation was inspected."]},{"id":"chronos-knn","date":"February–March 2026","url":"/projects/chronos-knn","visual":"forecast","why":"Investigate whether historical pattern retrieval and adapted foundation models help with decisions on stock time series.","question":"Does improving the prediction objective improve the behavior of the decision rule that consumes it?","hard":"Feature fitting must respect time. Input and target units must match. A ranking metric can improve while the position behavior becomes worse.","assumptions":"Historical windows and labels need appropriate time boundaries. Architecture-level causal filters alone do not certify every evaluation as leakage-free.","data":"Twenty-day stock windows, normalized raw-feature retrieval, and separate historical adaptation/ablation studies. Models including Chronos and Granite TTM are upstream.","failure":"A trend-clarity objective made nearly every forecast positive. The simulator required negative trend clarity to exit, so its exit mechanism stopped firing.","changed":"The work moved from neural embeddings to a raw-feature baseline, and later explored adaptation and persistence classification. The later head is not claimed to fix the earlier failure.","lessons":"Test the entire prediction-to-decision chain. A smooth output is not an economic objective.","unresolved":"Robustness across regimes, capital-constrained execution, and external or live use are not established by these studies.","attribution":"Donald-attributed implementation and experiment history. Foundation models and their research are credited as upstream work.","name":"Time-series forecasting & retrieval","title":"Time-series forecasting & retrieval","description":"Time-series foundation models meet nearest-neighbor retrieval, causal evaluation, and ablations that expose prediction-to-decision failures.","category":"Machine Learning & Model Research","technologies":["PyTorch","Qdrant","Time series","Walk-forward evaluation","Chronos","Kronos","Granite TTM","LoRA"],"roles":["AI / ML","Quant Research","Research"],"contribution":"Donald built raw-feature retrieval baselines, model-adaptation experiments, causal PCA, and persistence-head training around upstream time-series models. The related Kronos RAG workspace is a separate local study.","result":"The research framework connects retrieval, forecasting, and decision-rule evaluation. Ablations exposed an objective that disabled simulated exits, while regime-selected retraining did not outperform its rolling baseline in the documented study.","evidence":["chronos-architecture","chronos-failure","chronos-regime","chronos-knn-catalog-1","chronos-knn-catalog-2","chronos-knn-catalog-3","chronos-knn-catalog-4"],"field":"machine-learning","rank":1,"rankReason":"The deepest experimental thread: model adaptation, retrieval, ablations, and useful negative results.","folders":["chronos_knn","A_Share/kronos_rag"],"kind":"built","fullStory":true,"limitations":["Upstream foundation models are not Donald inventions. No live-performance claim. Architecture time filters alone do not certify full label-availability correctness; simulator leverage and calibration caveats preclude promoting headline returns. Later persistence heads are not proven to fix the earlier failure."]},{"id":"momentum-transformer-adaptation","date":"Project record · Sep 2026","url":"/projects/momentum-transformer-adaptation","visual":"schema","why":"An extensive adaptation of the upstream Momentum Transformer research, including a PyTorch implementation and portfolio/execution reinforcement-learning layers.","question":"","hard":"Attention/LSTM financial sequence modeling, direct risk-objective optimization, regime features, portfolio-aware action projection, PPO/SAC training, GAE, observation normalization, transaction-cost accounting and execution adapters.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Original papers and their reported results belong to upstream authors. Names such as production execution layer denote code organization, not verified deployed profitability. No model or trading tests rerun in this audit.","attribution":"Donald extended Kieran Wood’s Momentum Transformer research with PyTorch modeling, execution environments, trade-delta policies, portfolio-aware agents, PPO corrections, and replay visualization.","contribution":"Donald extended Kieran Wood’s Momentum Transformer research with PyTorch modeling, execution environments, trade-delta policies, portfolio-aware agents, PPO corrections, and replay visualization.","result":"Portfolio and execution-agent interfaces support threshold and reinforcement-learning policies. The work also corrected clipped-action log-probabilities and duplicate transaction penalties in the simulation/training path.","name":"Momentum Transformer & execution learning","title":"Momentum Transformer & execution learning","description":"An extensive adaptation of the upstream Momentum Transformer research, including a PyTorch implementation and portfolio/execution reinforcement-learning layers.","category":"Machine Learning & Model Research","technologies":["PyTorch","Temporal attention","LSTM","PPO","SAC","Execution simulation"],"roles":["AI / ML","Research","Quant Research"],"evidence":["momentum-transformer-adaptation-catalog-1","momentum-transformer-adaptation-catalog-2","momentum-transformer-adaptation-catalog-3","momentum-transformer-adaptation-catalog-4"],"field":"machine-learning","rank":2,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["trading-momentum-transformer"],"kind":"adaptation","fullStory":false,"limitations":["Original papers and their reported results belong to upstream authors. Names such as production execution layer denote code organization, not verified deployed profitability. No model or trading tests rerun in this audit."]},{"id":"mini-torch","date":"Project record · Sep 2026","url":"/projects/mini-torch","visual":"schema","why":"A NumPy-only neural-network training library with dynamic computation graphs, reverse-mode differentiation and a small MNIST classifier.","question":"","hard":"Function/Context graph nodes; gradient accumulation along multiple paths; unbroadcasting; batched matmul derivatives; reductions and activation derivatives; numerically stabilized softmax cross-entropy; module/parameter discovery; dataset/batching; NumPy-only optimizer and train/test preprocessing.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"A CS446/ECE449 course study based on supplied scaffolding, not original autodiff research. No Git history or explicit candidate-named report was found; public phrasing should retain the coursework attribution. Do not claim generic PyTorch compatibility or production performance. No measured classification accuracy is claimed.","attribution":"Course implementations build dynamic computation graphs, broadcast-aware derivatives, topological backpropagation, parameterized layers, SGD, and a 784–256–128–10 classifier from NumPy.","contribution":"Course implementations build dynamic computation graphs, broadcast-aware derivatives, topological backpropagation, parameterized layers, SGD, and a 784–256–128–10 classifier from NumPy.","result":"An end-to-end teaching library spanning differentiation, batching, optimization, and MNIST training code, with supplied unit tests.","name":"mini_torch — autograd from NumPy","title":"mini_torch — autograd from NumPy","description":"A NumPy-only neural-network training library with dynamic computation graphs, reverse-mode differentiation and a small MNIST classifier.","category":"Machine Learning & Model Research","technologies":["Python","NumPy","Automatic differentiation","Neural networks","MNIST"],"roles":["AI / ML","Research","Quant Research"],"evidence":["mini-torch-catalog-1","mini-torch-catalog-2","mini-torch-catalog-3","mini-torch-catalog-4","mini-torch-catalog-5"],"field":"machine-learning","rank":3,"rankReason":"Strong first-principles ML implementation, distinct from wrapping a foundation-model API; ranking is within this assigned subset only.","folders":["uiuc/ece449/CS446_ECE449_SP2026_MP1"],"kind":"study","fullStory":false,"limitations":["A CS446/ECE449 course study based on supplied scaffolding, not original autodiff research.","No Git history or explicit candidate-named report was found; public phrasing should retain the coursework attribution.","Do not claim generic PyTorch compatibility or production performance.","No measured classification accuracy is claimed."]},{"id":"rl-portfolio-framework","date":"Project record · Sep 2026","url":"/projects/rl-portfolio-framework","visual":"schema","why":"Recurrent PPO, portfolio environments, and related regime and fundamental-overlay studies. Each experiment retains its own evaluation boundary.","question":"","hard":"Recurrent policies, PID-Lagrangian drawdown constraints, holdings/turnover-aware simulation, online versus oracle regimes, event-gated entry, reporting-lag alignment, fundamental transforms and controlled signal ablations.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"No production-readiness, profitability or live-computability certification. Reference-paper presence is not authorship or completed reproduction. Theoretical prototypes and deployed execution must remain separate.","attribution":"Donald implemented the recurrent-policy portfolio framework and its starter. Related local TET studies add logistic overlays and fundamentals pipelines; the jump-model paper is a separate reference.","contribution":"Donald implemented the recurrent-policy portfolio framework and its starter. Related local TET studies add logistic overlays and fundamentals pipelines; the jump-model paper is a separate reference.","result":"Portfolio environments combine PPO-PID drawdown constraints, turnover-aware simulation, regime modules, and diagnostics. The research explicitly separates causal online labels from smoothed oracle labels.","name":"Regime-aware reinforcement learning","title":"Regime-aware reinforcement learning","description":"Recurrent PPO, portfolio environments, and related regime and fundamental-overlay studies. Each experiment retains its own evaluation boundary.","category":"Machine Learning & Model Research","technologies":["PyTorch","Gymnasium","PPO","RecurrentPPO","Jump models","GMM","Streamlit","Python","Tushare","Logistic regression","Fundamental factors"],"roles":["AI / ML","Research","Quant Research"],"evidence":["rl-portfolio-framework-catalog-1","rl-portfolio-framework-catalog-2","rl-portfolio-framework-catalog-3","rl-portfolio-framework-catalog-4","rl-portfolio-framework-catalog-5","rl-portfolio-framework-catalog-6","rl-portfolio-framework-catalog-7","rl-portfolio-framework-catalog-8"],"field":"machine-learning","rank":4,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["rl_trader_framework","lstm_ppo_regime_starter","tet-jumpmodels-tushare","tet_fundamentals_tushare_a_share","jump_model_regime"],"kind":"built","fullStory":false,"limitations":["No production-readiness, profitability or live-computability certification. Reference-paper presence is not authorship or completed reproduction. Theoretical prototypes and deployed execution must remain separate."]},{"id":"xtrend","date":"November 2025","url":"/projects/xtrend","visual":"attention","why":"Reproduce and investigate an existing financial time-series approach, connecting paper equations to the actual tensors.","question":"If a key retrieves a similar market state, does the value include its observed outcome?","hard":"A tensor with the right shape can encode the wrong meaning. Paper alignment requires checking both data availability and the information flowing through attention.","assumptions":"Context outcomes must be observed by prediction time. Matching an architecture does not reproduce its reported financial results.","data":"Historical financial time series and context windows. No proprietary source dataset is redistributed on this site.","failure":"The initial key and value encoders both used conditions alone, omitting returns from the value representation.","changed":"The corrected value projection concatenates context features and observed returns. Normalization and training protocols also evolved.","lessons":"Trace the meaning of every tensor, not only its dimensions.","unresolved":"Exact paper reproduction and economic validity require separate evaluation.","attribution":"Reproduction and adaptation of upstream X-Trend research. No claim to authorship of the original paper.","name":"X-Trend reproduction","title":"X-Trend reproduction","description":"Two iterations of a few-shot forecasting reproduction, including a tensor-level correction to what attention values remember.","category":"Machine Learning & Model Research","technologies":["PyTorch","Cross-attention","LSTM","Expanding windows","Gaussian processes","Sparse jump models"],"roles":["AI / ML","Research","Quant Research"],"contribution":"Donald’s agent-assisted adaptation of X-Trend implements Q/K/V projections and corrects attention values to include observed context returns. Two iterations form one evolving reproduction of the upstream research.","result":"The revised PyTorch construction lets retrieved context carry both market conditions and their observed outcomes. Experiment notes also record numerical instability in change-point fitting and later segmentation changes; original-paper performance remains unverified.","evidence":["xtrend-values","xtrend-source","xtrend-catalog-1","xtrend-catalog-2","xtrend-catalog-3","xtrend-catalog-4","xtrend-catalog-5"],"field":"machine-learning","rank":5,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["XTREND","xtrend_revised"],"kind":"adaptation","fullStory":true,"limitations":["Do not repeat the early README claim of superior returns or commit subjects naming target Sharpe as achieved performance. Architecture reproduction, economic reproduction and live trading remain distinct."]},{"id":"generative-models","date":"Project record · Sep 2026","url":"/projects/generative-models","visual":"schema","why":"A sequence of course studies in latent-variable generation, DDPM training and classifier-free guidance, then diffusion-based image deblurring.","question":"","hard":"Convolutional VAE with KL weighting and generation-distribution comparison; time-conditioned U-Net diffusion from supplied scaffolding; two-pass classifier-free guidance; Tweedie reconstruction; self-adjoint blur likelihood correction; six guidance scales and four noise settings.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Coursework using supplied architectures and OpenAI guided-diffusion/pretrained weights; those upstream models are not Donald-authored. The DPS report is qualitative and based on its selected image setup, not broad image-restoration performance. No aggregate PSNR/SSIM or reproducibility benchmark was established. mp2 lowercase is a setup duplicate; group it with MP2 rather than count it as another project.","attribution":"Course notebooks implement VAE objectives, DDPM training and sampling, classifier-free guidance, and simplified diffusion posterior sampling. Donald’s deblurring report compares six guidance scales and four measurement-noise settings.","contribution":"Course notebooks implement VAE objectives, DDPM training and sampling, classifier-free guidance, and simplified diffusion posterior sampling. Donald’s deblurring report compares six guidance scales and four measurement-noise settings.","result":"Generated-image artifacts and a qualitative deblurring study show the tradeoff: insufficient guidance loses the input, excessive guidance introduces artifacts, and higher measurement noise degrades reconstruction.","name":"VAE, DDPM & guided diffusion","title":"VAE, DDPM & guided diffusion","description":"A sequence of course studies in latent-variable generation, DDPM training and classifier-free guidance, then diffusion-based image deblurring.","category":"Machine Learning & Model Research","technologies":["Python","PyTorch","VAE","DDPM","Classifier-free guidance","Diffusion posterior sampling","Jupyter"],"roles":["AI / ML","Research","Quant Research"],"evidence":["generative-models-catalog-1","generative-models-catalog-2","generative-models-catalog-3","generative-models-catalog-4","generative-models-catalog-5","generative-models-catalog-6"],"field":"machine-learning","rank":6,"rankReason":"Broader model understanding plus directly named experiment analysis; less original system construction than mini_torch.","folders":["uiuc/ece598/mp1","uiuc/ece598/MP2","uiuc/ece598/mp2","uiuc/ece598/MP3"],"kind":"study","fullStory":false,"limitations":["Coursework using supplied architectures and OpenAI guided-diffusion/pretrained weights; those upstream models are not Donald-authored.","The DPS report is qualitative and based on its selected image setup, not broad image-restoration performance.","No aggregate PSNR/SSIM or reproducibility benchmark was established.","mp2 lowercase is a setup duplicate; group it with MP2 rather than count it as another project."]},{"id":"adaptive-portfolio-learning","date":"Project record · Sep 2026","url":"/projects/adaptive-portfolio-learning","visual":"schema","why":"Related DoubleAdapt/StockMixer, DeePM, and imitation-learning experiments in adapting models to A-share portfolio decisions.","question":"","hard":"Causal support/query adaptation, monthly retraining, market-specific execution constraints, utility labels, feasible teachers, DAgger scaffolding, conformal lower bounds, reliability abstention and walk-forward tests.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Original models and publications remain upstream. The negative DeePM report has universe/survivorship caveats and should be presented as its dated experiment, not a universal impossibility result. No headline financial metrics are promoted.","attribution":"Donald adapted the upstream DeePM codebase for A-share research. Related local DoubleAdapt/StockMixer and imitation-learning implementations explore causal adaptation, feasible teachers, and reliability-based abstention.","contribution":"Donald adapted the upstream DeePM codebase for A-share research. Related local DoubleAdapt/StockMixer and imitation-learning implementations explore causal adaptation, feasible teachers, and reliability-based abstention.","result":"Modular pipelines support walk-forward portfolio experiments. The documented DeePM study rejected the tested long-only selection thesis and separated market exposure from active selection value.","name":"A-share adaptive portfolio learning","title":"A-share adaptive portfolio learning","description":"Related DoubleAdapt/StockMixer, DeePM, and imitation-learning experiments in adapting models to A-share portfolio decisions.","category":"Machine Learning & Model Research","technologies":["PyTorch","DoubleAdapt","StockMixer","DeePM","Imitation learning","Conformal calibration","LightGBM"],"roles":["AI / ML","Research","Quant Research"],"evidence":["adaptive-portfolio-learning-catalog-1","adaptive-portfolio-learning-catalog-2","adaptive-portfolio-learning-catalog-3","adaptive-portfolio-learning-catalog-4","adaptive-portfolio-learning-catalog-5"],"field":"machine-learning","rank":7,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/doubleadapt_stockmixer","A_Share/deepm","A_Share/FSD"],"kind":"adaptation","fullStory":false,"limitations":["Original models and publications remain upstream. The negative DeePM report has universe/survivorship caveats and should be presented as its dated experiment, not a universal impossibility result. No headline financial metrics are promoted."]},{"id":"project-genji","date":"Project record · Sep 2026","url":"/projects/project-genji","visual":"schema","why":"An industry-aware equity research framework whose concrete implementation centers on ingestion, feature/label contracts, purged cross-validation and supervised model training.","question":"","hard":"Cross-sectional winsorization, industry normalization, multi-horizon training, purged/embargoed temporal splits, deterministic model comparison and structured prediction/evaluation contracts.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Do not call all originally planned regime, QP portfolio, RL and backtest components complete. README plans and current implemented supervised-learning path must be distinguished.","attribution":"Donald implemented feature and label contracts, time-aware cross-validation, supervised training, out-of-fold aggregation, and model persistence for industry-aware equity research.","contribution":"Donald implemented feature and label contracts, time-aware cross-validation, supervised training, out-of-fold aggregation, and model persistence for industry-aware equity research.","result":"Ridge and XGBoost training share purged and embargoed temporal splits, fold metrics, and aggregated out-of-fold predictions for structured model comparison.","name":"Project Genji: temporal model evaluation","title":"Project Genji: temporal model evaluation","description":"An industry-aware equity research framework whose concrete implementation centers on ingestion, feature/label contracts, purged cross-validation and supervised model training.","category":"Machine Learning & Model Research","technologies":["Python","Qlib","XGBoost","Ridge regression","Purged cross-validation","Tushare"],"roles":["AI / ML","Research","Quant Research"],"evidence":["project-genji-catalog-1","project-genji-catalog-2","project-genji-catalog-3","project-genji-catalog-4"],"field":"machine-learning","rank":8,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["donald_trading_model"],"kind":"built","fullStory":false,"limitations":["Do not call all originally planned regime, QP portfolio, RL and backtest components complete. README plans and current implemented supervised-learning path must be distinguished."]},{"id":"symbolic-alpha-mining","date":"Project record · Sep 2026","url":"/projects/symbolic-alpha-mining","visual":"schema","why":"An adapted AlphaPROBE research workspace with local point-in-time feature infrastructure and alpha-allocation additions.","question":"","hard":"Symbolic expression trees, alpha pools, knowledge-guided search, GFlowNet/PPO search entrypoints, point-in-time feature access and allocation after discovery.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Original AlphaPROBE paper/model is not Donald’s publication. Working-tree additions establish local work, not validated alpha or exact sole authorship. No research outputs inspected.","attribution":"Local adaptations extend the upstream AlphaPROBE codebase with a point-in-time feature registry, dataset adapters, alpha-allocation modules, and targeted tests. The original algorithm and research belong to the paper authors.","contribution":"Local adaptations extend the upstream AlphaPROBE codebase with a point-in-time feature registry, dataset adapters, alpha-allocation modules, and targeted tests. The original algorithm and research belong to the paper authors.","result":"The workspace connects symbolic-alpha search to explicit feature-availability controls and allocation code, extending the upstream discovery workflow.","name":"Symbolic alpha mining and allocation","title":"Symbolic alpha mining and allocation","description":"An adapted AlphaPROBE research workspace with local point-in-time feature infrastructure and alpha-allocation additions.","category":"Machine Learning & Model Research","technologies":["Python","Symbolic expressions","GFlowNet","PPO","Point-in-time data"],"roles":["AI / ML","Research","Quant Research"],"evidence":["symbolic-alpha-mining-catalog-1","symbolic-alpha-mining-catalog-2","symbolic-alpha-mining-catalog-3"],"field":"machine-learning","rank":9,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/factorResearch/AlphaPROBE"],"kind":"adaptation","fullStory":false,"limitations":["Original AlphaPROBE paper/model is not Donald’s publication. Working-tree additions establish local work, not validated alpha or exact sole authorship. No research outputs inspected."]},{"id":"visual-alpha-research","date":"Project record · Sep 2026","url":"/projects/visual-alpha-research","visual":"schema","why":"Chart-image CNN/ViT, market-guided attention, and cross-sectional representation studies in one model-research family.","question":"","hard":"Reproducible chart rendering, visual-versus-numeric comparisons, market-conditioned adapters, characteristic attention, variable-universe masks, PIT alignment and portfolio construction under market constraints.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"No paper publication, successful SDF replication, model advantage or deployment is established. Planned visual fusion and directly executable market variants must not be conflated with finished evaluation.","attribution":"Local implementations and adaptations explore chart-image CNNs/ViTs, MASTER adapters, and Attention Factors. Each original paper and model remains credited to its authors.","contribution":"Local implementations and adaptations explore chart-image CNNs/ViTs, MASTER adapters, and Attention Factors. Each original paper and model remains credited to its authors.","result":"The family includes chart rendering, model and evaluation modules, and ViT inference/visualization. Attention Factors distinguishes the paper’s long/short reference from a long-tilt hedge construction.","name":"A-share representation and attention-model adaptations","title":"A-share representation and attention-model adaptations","description":"Chart-image CNN/ViT, market-guided attention, and cross-sectional representation studies in one model-research family.","category":"Machine Learning & Model Research","technologies":["PyTorch","CNN","Vision Transformer","MASTER","t-SNE","A-share data","Attention factors","Statistical arbitrage","Point-in-time panels"],"roles":["AI / ML","Research","Quant Research"],"evidence":["visual-alpha-research-catalog-1","visual-alpha-research-catalog-2","visual-alpha-research-catalog-3","visual-alpha-research-catalog-4","visual-alpha-research-catalog-5","visual-alpha-research-catalog-6","visual-alpha-research-catalog-7"],"field":"machine-learning","rank":10,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/factorResearch/CNN","A_Share/factorResearch/vit-sdf-lasso","A_Share/factorResearch/MASTER","A_Share/factorResearch/visualization","A_Share/attention_factors"],"kind":"adaptation","fullStory":false,"limitations":["No paper publication, successful SDF replication, model advantage or deployment is established. Planned visual fusion and directly executable market variants must not be conflated with finished evaluation."]},{"id":"rotation-learning","date":"Project record · Sep 2026","url":"/projects/rotation-learning","visual":"schema","why":"Graph, GNN, and HMM experiments investigate which stocks recover after a market leader’s limit-up streak ends.","question":"","hard":"Within-event ranking, matured-label graph updates, temporal encoders, filtered HMM posteriors, weighted Plackett-Luce choice, out-of-fold calibration, lower-confidence-bound expected value and abstention.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"No leaderboard or trading-performance claims. Production wording in local model docs does not establish deployment. Do not combine experiment variants into a single validated algorithm.","attribution":"Local research implementations compare a signed temporal graph, a dynamic latent GNN, and a Student-t HMM choice model, with separate data and evaluation contracts.","contribution":"Local research implementations compare a signed temporal graph, a dynamic latent GNN, and a Student-t HMM choice model, with separate data and evaluation contracts.","result":"Alternative ranking paths use matured labels, filtered states, and explicit feature-availability rules. The GNN path excludes present-day industry classifications because they are not historical point-in-time labels.","name":"Leader-follower rotation and event ranking","title":"Leader-follower rotation and event ranking","description":"Graph, GNN, and HMM experiments investigate which stocks recover after a market leader’s limit-up streak ends.","category":"Machine Learning & Model Research","technologies":["GNN","Student-t HMM","Temporal encoders","Plackett-Luce","Calibration","Event ranking"],"roles":["AI / ML","Research","Quant Research"],"evidence":["rotation-learning-catalog-1","rotation-learning-catalog-2","rotation-learning-catalog-3","rotation-learning-catalog-4","rotation-learning-catalog-5"],"field":"machine-learning","rank":11,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/Rotation","A_Share/Rotation_hmm_v14_bg"],"kind":"built","fullStory":false,"limitations":["No leaderboard or trading-performance claims. Production wording in local model docs does not establish deployment. Do not combine experiment variants into a single validated algorithm."]},{"id":"cnn-ablation-harness","date":"Project record · Sep 2026","url":"/projects/cnn-ablation-harness","visual":"schema","why":"A modular reproduction of the PyTorch CIFAR-10 tutorial that turns seed, augmentation, epoch and channel-width changes into selectable experiment configurations.","question":"","hard":"Reproducibility controls; configurable affine transforms; architecture-width variants; train/evaluation loop; machine-readable result output; modular ablation harness.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Tutorial-based course work, not a new CNN architecture. No independent candidate attribution or fresh test run. The course variation applies augmentation to both training and test transforms, so comparisons need that evaluation caveat.","attribution":"Local modules separate configuration, seeding, data transforms, model, training and evaluation, with CLI experiment flags and lightweight wiring/smoke tests.","contribution":"Local modules separate configuration, seeding, data transforms, model, training and evaluation, with CLI experiment flags and lightweight wiring/smoke tests.","result":"A configurable CIFAR-10 experiment harness with separated training/evaluation modules, machine-readable results, and smoke-test source. README sample scores are illustrative rather than measured outcomes.","name":"CIFAR-10 experiment harness","title":"CIFAR-10 experiment harness","description":"A modular reproduction of the PyTorch CIFAR-10 tutorial that turns seed, augmentation, epoch and channel-width changes into selectable experiment configurations.","category":"Machine Learning & Model Research","technologies":["Python","PyTorch","CIFAR-10","Experiment configuration"],"roles":["AI / ML","Research","Quant Research"],"evidence":["cnn-ablation-harness-catalog-1","cnn-ablation-harness-catalog-2","cnn-ablation-harness-catalog-3","cnn-ablation-harness-catalog-4","cnn-ablation-harness-catalog-5"],"field":"machine-learning","rank":12,"rankReason":"Useful evidence of methodical experimentation, but more elementary and tutorial-bound than the other ML work.","folders":["uiuc/cs441"],"kind":"study","fullStory":false,"limitations":["Tutorial-based course work, not a new CNN architecture.","No independent candidate attribution or fresh test run.","The course variation applies augmentation to both training and test transforms, so comparisons need that evaluation caveat."]},{"id":"upstream-ai-reference-library","date":"Project record · Sep 2026","url":"/projects/upstream-ai-reference-library","visual":"schema","why":"CAMEF and the Kalshi AI bot: upstream codebases retained as reference foundations, with original authorship clearly distinguished.","question":"","hard":"Reference topics are multimodal/counterfactual forecasting and LLM-based market-agent systems; these describe upstream scope only.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Neither upstream publication, KDD acceptance, original model, codebase, release, user count nor performance is Donald’s. Folder presence alone does not establish completed study or endorsement.","attribution":"Reference study of upstream CAMEF and Ryan Frigo’s Kalshi AI bot. These clean checkouts are original authors’ work, with no local implementation contribution claimed.","contribution":"Reference study of upstream CAMEF and Ryan Frigo’s Kalshi AI bot. These clean checkouts are original authors’ work, with no local implementation contribution claimed.","result":"Reference implementations for multimodal forecasting and LLM market agents, cataloged separately from the local World Cup announcer-mention study.","name":"Forecasting & market-agent reference library","title":"Forecasting & market-agent reference library","description":"CAMEF and the Kalshi AI bot: upstream codebases retained as reference foundations, with original authorship clearly distinguished.","category":"Machine Learning & Model Research","technologies":["Multimodal learning","Causal forecasting","Counterfactual augmentation","Python","LLM applications","Prediction markets"],"roles":["AI / ML","Research","Quant Research"],"evidence":["upstream-ai-reference-library-catalog-1","upstream-ai-reference-library-catalog-2","upstream-ai-reference-library-catalog-3","upstream-ai-reference-library-catalog-4","upstream-ai-reference-library-catalog-5"],"field":"machine-learning","rank":13,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["CAMEF","kalshi"],"kind":"reference","fullStory":false,"limitations":["Neither upstream publication, KDD acceptance, original model, codebase, release, user count nor performance is Donald’s. Folder presence alone does not establish completed study or endorsement."]},{"id":"local-qwen-inference","date":"Project record · Sep 2026","url":"/projects/local-qwen-inference","visual":"schema","why":"Self-hosted Qwen workflows with controlled request budgets, reproducible decoding, structured decisions, and explicit inference-failure handling.","question":"","hard":"Durable call reservations use file locks and fsync, including failed requests. Decoding parameters are pinned, model/runtime metadata is captured, JSON-schema output is requested, and completion budgets scale with target count. The client explicitly disables reasoning for concise structured outputs after discovering null-content/truncation behavior.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"This is a capability collection across projects, not a separate published model or original inference engine. No throughput benchmark, SLA, model-weight provenance audit, external-customer deployment, or training claim is established. The autoresearch client is present in the working tree but has no path-specific committed attribution in the inspected history.","attribution":"Application-side controls wrap upstream Qwen/vLLM with durable request budgets, fixed decoding settings, runtime metadata, and structured outputs. Donald also contributed private-serving configuration in the event-ending project.","contribution":"Application-side controls wrap upstream Qwen/vLLM with durable request budgets, fixed decoding settings, runtime metadata, and structured outputs. Donald also contributed private-serving configuration in the event-ending project.","result":"Local Qwen generation was exercised successfully. The client handles output truncation and distinguishes transport, HTTP, parsing, and completion failures; this is inference integration, not a new model or serving engine.","name":"Local Qwen inference","title":"Local Qwen inference","description":"Self-hosted Qwen workflows with controlled request budgets, reproducible decoding, structured decisions, and explicit inference-failure handling.","category":"Local LLM Inference & Agents","technologies":["Qwen","vLLM","OpenAI-compatible HTTP","JSON Schema","Python","fcntl","fsync","Quantization","PyTorch"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["local-qwen-inference-catalog-1","local-qwen-inference-catalog-2","local-qwen-inference-catalog-3","local-qwen-inference-catalog-4"],"field":"local-llms","rank":1,"rankReason":"Concrete inference engineering across projects: budgets, structured outputs, runtime controls, and recovery.","folders":["kalshibot-membrane/autoresearch/lab","market-ending","aisop/packages/aisop/adjudication"],"kind":"adaptation","fullStory":false,"limitations":["This is a capability collection across projects, not a separate published model or original inference engine.","No throughput benchmark, SLA, model-weight provenance audit, external-customer deployment, or training claim is established.","The autoresearch client is present in the working tree but has no path-specific committed attribution in the inspected history."]},{"id":"zhengmind-yagni","date":"Project record · Sep 2026","url":"/projects/zhengmind-yagni","visual":"schema","why":"An AI-work coordinator built around Git, markdown work nodes, a small state machine, and a CLI that owns mutations.","question":"","hard":"Dependency-ready claims use O_EXCL file locks. The core separates work-tree structure, transition rules, events, and persistence. The web conductor invokes planner/worker/reviewer processes while the CLI remains the state-mutation boundary; lesson memory and Git-head/live-deploy comparison make unfinished work visible.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"README statements of about 1.6k lines and a purely read-only dashboard are stale: current source has more code and POST endpoints that start agent loops. No production, multi-tenant, distributed coordination, or external-adoption claim is established. The architectural critique is project documentation, not a verified first-person quotation. The coordination engine is local; diagnostic agent calls use Claude CLI and are not local-model inference.","attribution":"Donald simplified the state machine, removed unused machinery, added reusable lessons, and made agent-loop progress observable. The implementation keeps state changes behind a CLI and uses exclusive claims for ready work.","contribution":"Donald simplified the state machine, removed unused machinery, added reusable lessons, and made agent-loop progress observable. The implementation keeps state changes behind a CLI and uses exclusive claims for ready work.","result":"A dependency-free Go coordinator for planner, worker, and reviewer processes, with durable work nodes and transition/claim tests. Work records show the workflow applied to the YouTube Knowledge project.","name":"ZhengMindYAGNI — agent orchestration","title":"ZhengMindYAGNI — agent orchestration","description":"An AI-work coordinator built around Git, markdown work nodes, a small state machine, and a CLI that owns mutations.","category":"Local LLM Inference & Agents","technologies":["Go","Git","Markdown","JSONL","File locks","CLI","HTTP","Claude CLI"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["zhengmind-yagni-catalog-1","zhengmind-yagni-catalog-2","zhengmind-yagni-catalog-3","zhengmind-yagni-catalog-4","zhengmind-yagni-catalog-5","zhengmind-yagni-catalog-6"],"field":"local-llms","rank":2,"rankReason":"Lead agent orchestration: strongest candidate-attributed system and a concrete simplification story grounded in source and history.","folders":["ZhengMindYAGNI"],"kind":"built","fullStory":false,"limitations":["README statements of about 1.6k lines and a purely read-only dashboard are stale: current source has more code and POST endpoints that start agent loops.","No production, multi-tenant, distributed coordination, or external-adoption claim is established.","The architectural critique is project documentation, not a verified first-person quotation.","The coordination engine is local; diagnostic agent calls use Claude CLI and are not local-model inference."]},{"id":"vta-qwen-adaptation","date":"Project record · Sep 2026","url":"/projects/vta-qwen-adaptation","visual":"schema","why":"A local adaptation of an upstream financial time-series LLM pipeline to Qwen3.5, spanning inference compatibility and staged GRPO/LoRA/SFT training code.","question":"","hard":"Typed chat content and processor normalization handle Qwen3.5 multimodal architecture conventions. Loader/generation paths distinguish FastModel from FastLanguageModel and disable the vLLM fast path for this architecture. Pipeline stages combine reward training, rejection sampling, supervised fine-tuning, and evaluation.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Upstream algorithm, training pipeline, and paper authorship must be credited separately. README marketing about improved forecasting is not a verified result of this adaptation. Local compatibility branches are environment-specific and not a universal statement that Qwen3.5 cannot run on vLLM. No predictive advantage, live trading value, or training completion is established.","attribution":"Local changes adapt the upstream chen-jan pipeline to Qwen3.5 chat and processor conventions, distinguish model-loading paths, and add an environment-specific dependency profile. The original algorithms and training workflow remain upstream work.","contribution":"Local changes adapt the upstream chen-jan pipeline to Qwen3.5 chat and processor conventions, distinguish model-loading paths, and add an environment-specific dependency profile. The original algorithms and training workflow remain upstream work.","result":"Compatibility code spans generation and staged GRPO/LoRA/SFT workflows. A completed training run or forecasting improvement has not been established.","name":"VTA — local Qwen adaptation","title":"VTA — local Qwen adaptation","description":"A local adaptation of an upstream financial time-series LLM pipeline to Qwen3.5, spanning inference compatibility and staged GRPO/LoRA/SFT training code.","category":"Local LLM Inference & Agents","technologies":["Qwen3.5","Unsloth","PyTorch","Transformers","TRL","LoRA","GRPO","SFT","CUDA","Python"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["vta-qwen-adaptation-catalog-1","vta-qwen-adaptation-catalog-2","vta-qwen-adaptation-catalog-3","vta-qwen-adaptation-catalog-4","vta-qwen-adaptation-catalog-5"],"field":"local-llms","rank":3,"rankReason":"Important local-model breadth: nontrivial Qwen3.5 compatibility work; no completed-training or forecasting-performance proof.","folders":["A_Share/VTA"],"kind":"adaptation","fullStory":false,"limitations":["Upstream algorithm, training pipeline, and paper authorship must be credited separately.","README marketing about improved forecasting is not a verified result of this adaptation.","Local compatibility branches are environment-specific and not a universal statement that Qwen3.5 cannot run on vLLM.","No predictive advantage, live trading value, or training completion is established."]},{"id":"qwen-market-ending","date":"Project record · Sep 2026","url":"/projects/qwen-market-ending","visual":"schema","why":"A local-Qwen decision pipeline asks whether the real-world event behind a prediction market is likely to end within a chosen horizon.","question":"","hard":"The pipeline groups related markets around a real-world event, hides the target moneyline, separates official feed facts from market-price evidence, and supports monotonic multi-horizon outputs. Uncertainty is an explicit downstream risk state rather than forced certainty.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Small opportunistic sample with mostly straightforward positives; no robust generalization or economic-value claim. The documented ground truth uses market close time, which should not be assumed identical to real event end without further validation. Cross-exchange game identity is not fully unified.","attribution":"A shared implementation combines exchange facts, companion-market context, local-model calls, and comparison with a rule baseline. Donald contributed configurable private serving; the source also records jd contributions.","contribution":"A shared implementation combines exchange facts, companion-market context, local-model calls, and comparison with a rule baseline. Donald contributed configurable private serving; the source also records jd contributions.","result":"A dated 25-game exploratory comparison reported 22 judged-correct decisions and three unsure. The small sample and market-close labels do not establish general accuracy or economic value.","name":"Event-ending inference with Qwen","title":"Event-ending inference with Qwen","description":"A local-Qwen decision pipeline asks whether the real-world event behind a prediction market is likely to end within a chosen horizon.","category":"Local LLM Inference & Agents","technologies":["Python standard library","Qwen","vLLM","HTTP APIs","JSON","Concurrent futures","Evaluation"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["qwen-market-ending-catalog-1","qwen-market-ending-catalog-2","qwen-market-ending-catalog-3","qwen-market-ending-catalog-4"],"field":"local-llms","rank":4,"rankReason":"High relevance to local LLMs and markets: inspectable live-fact pipeline and exploratory comparison; mixed authorship and small-sample limits constrain outcome claims.","folders":["market-ending"],"kind":"built","fullStory":false,"limitations":["Small opportunistic sample with mostly straightforward positives; no robust generalization or economic-value claim.","The documented ground truth uses market close time, which should not be assumed identical to real event end without further validation.","Cross-exchange game identity is not fully unified."]},{"id":"semantic-research-agents","date":"Project record · Sep 2026","url":"/projects/semantic-research-agents","visual":"schema","why":"A-share research workflows that turn textual evidence into structured long-only decisions, with accounting and governance checks.","question":"","hard":"Structured model outputs, point-in-time evidence, research orchestration, confirmation versus veto logic, and constrained portfolio construction.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Original frameworks and papers remain upstream. ATLAS subscriber/revenue claims and VTA state-of-the-art claims must never be attributed to Donald. LLM reasoning quality, training success, financial performance and adoption are unverified.","attribution":"Donald-attributed commits extend the upstream TradingAgents and ATLAS frameworks with specialized A-share research workflows. A related standalone semantic engine implements confirmation signals, vetoes and constrained portfolios.","contribution":"Donald-attributed commits extend the upstream TradingAgents and ATLAS frameworks with specialized A-share research workflows. A related standalone semantic engine implements confirmation signals, vetoes and constrained portfolios.","result":"A set of specialized research agents, structured decision modules, and dashboards. Confirmation signals, vetoes, and portfolio constraints make the proposed decisions inspectable; the experiments retain separate validation boundaries.","name":"Semantic research agents","title":"Semantic research agents","description":"A-share research workflows that turn textual evidence into structured long-only decisions, with accounting and governance checks.","category":"Local LLM Inference & Agents","technologies":["Python","LLM agents","Structured outputs","TradingAgents","ATLAS","Tushare"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["semantic-research-agents-catalog-1","semantic-research-agents-catalog-2","semantic-research-agents-catalog-3","semantic-research-agents-catalog-4"],"field":"local-llms","rank":5,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/ashare_semantic_long","A_Share/tradingagents","A_Share/atlas"],"kind":"adaptation","fullStory":false,"limitations":["Original frameworks and papers remain upstream. ATLAS subscriber/revenue claims and VTA state-of-the-art claims must never be attributed to Donald. LLM reasoning quality, training success, financial performance and adoption are unverified."]},{"id":"agent-context-blueprint","date":"Project record · Sep 2026","url":"/projects/agent-context-blueprint","visual":"schema","why":"A reusable agent workflow that separates planning, interface skeletons, implementation, and integration through typed patch artifacts and deterministic hooks.","question":"","hard":"Patch packages carry tickets, diffs, tests, and review context. Hooks guard direct writes, build context packs, enforce branch/prompt policies, invoke Codex review, and drive a test/integration pipeline. Context windows are separated by role and token budgets are explicit.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"A workflow/template is not a hosted product or evidence of autonomous reliability at scale. Some starter assets and examples are scaffolding, not deployed application work. Hooks execute commands and require project-specific review; no runtime or security validation was performed.","attribution":"Donald built hook-driven integration, branch policy, durable review outputs, and documentation around typed handoffs between planning, implementation, and review roles.","contribution":"Donald built hook-driven integration, branch policy, durable review outputs, and documentation around typed handoffs between planning, implementation, and review roles.","result":"Patch packages carry tickets, diffs, tests, and review context. Deterministic hooks construct context packs, enforce workflow boundaries, and drive review and integration.","name":"Typed handoffs for coding agents","title":"Typed handoffs for coding agents","description":"A reusable agent workflow that separates planning, interface skeletons, implementation, and integration through typed patch artifacts and deterministic hooks.","category":"Local LLM Inference & Agents","technologies":["Python","Shell","JSON Schema","Git","Claude Code hooks","Codex","Agent orchestration"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["agent-context-blueprint-catalog-1","agent-context-blueprint-catalog-2","agent-context-blueprint-catalog-3","agent-context-blueprint-catalog-4"],"field":"local-llms","rank":6,"rankReason":"Useful engineering-evolution evidence: typed handoffs and deterministic integration hooks precede the later conductor.","folders":["claude-context-blueprint-template"],"kind":"built","fullStory":false,"limitations":["A workflow/template is not a hosted product or evidence of autonomous reliability at scale.","Some starter assets and examples are scaffolding, not deployed application work.","Hooks execute commands and require project-specific review; no runtime or security validation was performed."]},{"id":"claude-code-autopilot","date":"Project record · Sep 2026","url":"/projects/claude-code-autopilot","visual":"schema","why":"An installation and activation layer combining upstream coding skills, specialized agents, and hooks into a reusable project toolkit.","question":"","hard":"Prompt hooks select relevant skills, session hooks restore context, installation backs up existing files and installs hook dependencies, and review workflow was revised away from a blocking hook toward a skill-based approach.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Upstream skill counts and production-tested wording belong to the integrated ecosystems, not independent proof of Donald deployment outcomes. No quantified productivity benefit or broad installation compatibility was established.","attribution":"Donald built the installer and refined session-start and review behavior around upstream obra/superpowers and diet103 skills. Those integrated skills and agents remain credited to their authors.","contribution":"Donald built the installer and refined session-start and review behavior around upstream obra/superpowers and diet103 skills. Those integrated skills and agents remain credited to their authors.","result":"A reusable setup script backs up existing files, installs hook dependencies, and activates context and review workflows for coding projects.","name":"Claude Code Autopilot","title":"Claude Code Autopilot","description":"An installation and activation layer combining upstream coding skills, specialized agents, and hooks into a reusable project toolkit.","category":"Local LLM Inference & Agents","technologies":["Shell","Python","TypeScript","Claude Code","Hooks","Codex","Skills"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["claude-code-autopilot-catalog-1","claude-code-autopilot-catalog-2","claude-code-autopilot-catalog-3","claude-code-autopilot-catalog-4"],"field":"local-llms","rank":7,"rankReason":"Useful reusable tooling adaptation; rank below original systems because most skills and agents come from credited upstream projects.","folders":["claude-code-autopilot"],"kind":"adaptation","fullStory":false,"limitations":["Upstream skill counts and production-tested wording belong to the integrated ecosystems, not independent proof of Donald deployment outcomes.","No quantified productivity benefit or broad installation compatibility was established."]},{"id":"zhengmind-foundation","date":"Project record · Sep 2026","url":"/projects/zhengmind-foundation","visual":"schema","why":"A provider-neutral task-control architecture studied alongside the later ZhengMindYAGNI implementation.","question":"","hard":"Task trees, dependency-aware claims, scoped API tokens, prompt snapshots, sanitized audit events, proposal/SCM seams, and explicit parent integration express a sophisticated agent-governance model.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Not a production deployment; real external adapters and LLM execution are not established. Some domain concepts are future plans rather than implemented runtime support. Historical test counts in README were not rerun.","attribution":"An architectural study of the ZhengMind prototype credited to yq77zs73/Zheng Shen. It is separate from Donald’s independently attributed ZhengMindYAGNI implementation.","contribution":"An architectural study of the ZhengMind prototype credited to yq77zs73/Zheng Shen. It is separate from Donald’s independently attributed ZhengMindYAGNI implementation.","result":"The reference prototype explores task trees, dependency-aware claims, prompt snapshots, and auditable state changes. Its external database, GitHub-write, and LLM adapters remain unconnected.","name":"ZhengMind architecture study","title":"ZhengMind architecture study","description":"A provider-neutral task-control architecture studied alongside the later ZhengMindYAGNI implementation.","category":"Local LLM Inference & Agents","technologies":["Python","FastAPI","SQLite","Jinja2","Task graphs","Scoped tokens","Audit trails"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["zhengmind-foundation-catalog-1","zhengmind-foundation-catalog-2","zhengmind-foundation-catalog-3","zhengmind-foundation-catalog-4","zhengmind-foundation-catalog-5"],"field":"local-llms","rank":8,"rankReason":"Retain as an architectural study connected to YAGNI; separate authorship and mock integrations rule out using it as Donald built-work proof.","folders":["ZhengMind"],"kind":"study","fullStory":false,"limitations":["Not a production deployment; real external adapters and LLM execution are not established.","Some domain concepts are future plans rather than implemented runtime support.","Historical test counts in README were not rerun."]},{"id":"serena-foundation","date":"Project record · Sep 2026","url":"/projects/serena-foundation","visual":"schema","why":"An upstream reference for symbol-aware code retrieval and editing through language servers and MCP.","question":"","hard":"Upstream Serena connects language-server symbol information to agent tools such as find_symbol, find_referencing_symbols, and insert_after_symbol; MCP decouples the tools from a particular model/client.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Workspace presence does not prove extensive use, deployment, or original work. All capability, adoption, benchmark, and productivity claims belong to upstream unless separately substantiated.","attribution":"Reference study of the upstream Serena toolkit; no original implementation or local integration is claimed.","contribution":"Reference study of the upstream Serena toolkit; no original implementation or local integration is claimed.","result":"A concrete reference for connecting symbol definitions and references to model-facing editing tools.","name":"Serena semantic tooling reference","title":"Serena semantic tooling reference","description":"An upstream reference for symbol-aware code retrieval and editing through language servers and MCP.","category":"Local LLM Inference & Agents","technologies":["Python","MCP","Language Server Protocol","Semantic code tools"],"roles":["LLM Infrastructure","AI / ML","Software Engineering"],"evidence":["serena-foundation-catalog-1","serena-foundation-catalog-2"],"field":"local-llms","rank":9,"rankReason":"Retain in foundations only: clean upstream checkout establishes a reference, not candidate contribution.","folders":["serena"],"kind":"reference","fullStory":false,"limitations":["Workspace presence does not prove extensive use, deployment, or original work.","All capability, adoption, benchmark, and productivity claims belong to upstream unless separately substantiated."]},{"id":"isafe-sop","date":"Project record · Sep 2026","url":"/projects/isafe-sop","visual":"schema","why":"An assembly-inspection prototype that tracks procedure steps, flags skips and ordering errors, and attaches video evidence for review.","question":"","hard":"SOP steps form a DAG; sustained observations debounce noisy detections. Mandatory-ancestor precedence distinguishes alternate paths, skipped steps, and late wrong-order completions. Soft-DTW is downsampled and normalized to bound per-frame work. Evidence clips feed a second-opinion adjudicator, with a hash-chained audit ledger and SQLite/Postgres seams.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Mock CPU backends are the default. Real GPU inference and industrial deployment were not demonstrated by this audit. The current natural-language SOP compiler is a deterministic keyword parser; its LLM-backed compiler is a TODO seam despite broader product-language claims. Qwen2.5-VL 3B uses an opt-in NF4 4-bit adapter with visual-token bounds; parser tests do not establish real model accuracy. Field-signal adapters are stubs; production safety or defect-reduction outcomes are not established.","attribution":"Donald built the Python backend, integrated perception and adjudication components, repaired conformance logic, and iterated the review interface. Perception and vision-language models are upstream integrations.","contribution":"Donald built the Python backend, integrated perception and adjudication components, repaired conformance logic, and iterated the review interface. Perception and vision-language models are upstream integrations.","result":"The prototype connects few-shot registration, step conformance, video evidence, adjudication, persistence, and a review console. Its documented mock-perception acceptance scenario deliberately skips assembly step C to exercise procedure checking.","name":"ISAFE / WITNESS — SOP verification","title":"ISAFE / WITNESS — SOP verification","description":"An assembly-inspection prototype that tracks procedure steps, flags skips and ordering errors, and attaches video evidence for review.","category":"Software Systems & Automation","technologies":["Python","FastAPI","Pydantic","NumPy","Soft-DTW","DAG/FSM","Qwen2.5-VL","PyTorch","bitsandbytes","DINOv2","FAISS","SQLite","PostgreSQL","MQTT","WebSocket"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["isafe-sop-catalog-1","isafe-sop-catalog-2","isafe-sop-catalog-3","isafe-sop-catalog-4","isafe-sop-catalog-5","isafe-sop-catalog-6"],"field":"software-systems","rank":2,"rankReason":"An end-to-end workflow connecting perception, procedure logic, evidence review, storage, and an operator interface.","folders":["aisop"],"kind":"built","fullStory":false,"limitations":["Mock CPU backends are the default. Real GPU inference and industrial deployment were not demonstrated by this audit.","The current natural-language SOP compiler is a deterministic keyword parser; its LLM-backed compiler is a TODO seam despite broader product-language claims.","Qwen2.5-VL 3B uses an opt-in NF4 4-bit adapter with visual-token bounds; parser tests do not establish real model accuracy.","Field-signal adapters are stubs; production safety or defect-reduction outcomes are not established."]},{"id":"cpp-goroutine","date":"Project record · Sep 2026","url":"/projects/cpp-goroutine","visual":"schema","why":"Cooperative scheduling, epoll-backed I/O, timers, and synchronization in C++17. Waiting tasks yield instead of blocking a worker.","question":"","hard":"ucontext/makecontext/swapcontext scheduling across worker threads; mmap/mprotect stack guards; epoll and eventfd wakeups; deadline min-heap; read/write/recv/send/connect/accept/sleep hooks resolved with dlsym; user O_NONBLOCK tracking; parking handshake to avoid resuming before a context is saved; yielding mutexes and wait groups.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Experimental Linux-specific runtime, not a Go implementation or production concurrency guarantee. Examples do not establish scheduler race-freedom, cancellation correctness, stress reliability or benchmark throughput. No independent authorship/date history found. Throughput and latency have not been benchmarked.","attribution":"Implements a C++17 scheduler with guard-page stacks, epoll-backed I/O, timers, libc hooks, joinable tasks, and synchronization that yields cooperatively.","contribution":"Implements a C++17 scheduler with guard-page stacks, epoll-backed I/O, timers, libc hooks, joinable tasks, and synchronization that yields cooperatively.","result":"A library with CMake targets and examples for scheduling, timers, and a loopback echo server.","name":"A Go-style runtime in C++","title":"A Go-style runtime in C++","description":"Cooperative scheduling, epoll-backed I/O, timers, and synchronization in C++17. Waiting tasks yield instead of blocking a worker.","category":"Software Systems & Automation","technologies":["C++17","Linux","ucontext","epoll","eventfd","pthread","CMake"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["cpp-goroutine-catalog-1","cpp-goroutine-catalog-2","cpp-goroutine-catalog-3","cpp-goroutine-catalog-4","cpp-goroutine-catalog-5","cpp-goroutine-catalog-6","cpp-goroutine-catalog-7"],"field":"software-systems","rank":3,"rankReason":"Most distinctive omitted systems artifact: low-level implementation depth is directly visible, although attribution and runtime validation are weaker.","folders":["go-rountine"],"kind":"built","fullStory":false,"limitations":["Experimental Linux-specific runtime, not a Go implementation or production concurrency guarantee.","Examples do not establish scheduler race-freedom, cancellation correctness, stress reliability or benchmark throughput.","No independent authorship/date history found.","Throughput and latency have not been benchmarked."]},{"id":"applypilot-adaptation","date":"Project record · Sep 2026","url":"/projects/applypilot-adaptation","visual":"schema","why":"A local ApplyPilot adaptation adding a Claude CLI provider and crash-resilient parallel scoring.","question":"","hard":"The CLI adapter passes prompts through stdin, selects a model, avoids session persistence, isolates settings/CWD, and retries failures. Scoring uses ThreadPoolExecutor and periodically commits completed batches so a later crash loses less finished work.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"No candidate-authored committed provenance or executed test result for these edits was established. CLI invocation is a cloud-model integration, not local inference; upstream local-provider support is not a new candidate invention. README application counts are upstream promotional claims and must not be attributed to Donald. No application records, résumé artifacts, scraped job records, or candidate profile data were inspected.","attribution":"Extends Pickle-Pixel/Henry Muhiar’s ApplyPilot with an isolated Claude CLI adapter, retries, parallel scoring, and incremental SQLite commits. The underlying application agent is upstream work.","contribution":"Extends Pickle-Pixel/Henry Muhiar’s ApplyPilot with an isolated Claude CLI adapter, retries, parallel scoring, and incremental SQLite commits. The underlying application agent is upstream work.","result":"Completed scoring batches are committed incrementally so a later failure loses less finished work. The model adapter avoids persistent sessions and isolates its settings and working directory.","name":"ApplyPilot workflow adaptation","title":"ApplyPilot workflow adaptation","description":"A local ApplyPilot adaptation adding a Claude CLI provider and crash-resilient parallel scoring.","category":"Software Systems & Automation","technologies":["Python","httpx","Claude CLI","ThreadPoolExecutor","SQLite","Subprocess"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["applypilot-adaptation-catalog-1","applypilot-adaptation-catalog-2","applypilot-adaptation-catalog-3","applypilot-adaptation-catalog-4"],"field":"software-systems","rank":4,"rankReason":"Keep scoped adaptation visible because there are real provider, concurrency, and persistence edits; core application is upstream and modifications lack committed attribution.","folders":["applypilot"],"kind":"adaptation","fullStory":false,"limitations":["No candidate-authored committed provenance or executed test result for these edits was established.","CLI invocation is a cloud-model integration, not local inference; upstream local-provider support is not a new candidate invention.","README application counts are upstream promotional claims and must not be attributed to Donald.","No application records, résumé artifacts, scraped job records, or candidate profile data were inspected."]},{"id":"invoice-authentication","date":"Project record · Sep 2026","url":"/projects/invoice-authentication","visual":"schema","why":"Alternative native-authentication implementations inside a supplied React/GraphQL invoice application, with JWT access tokens, rotating refresh sessions and cookie-based browser integration.","question":"","hard":"Prisma-backed refresh records; rotation transactions; HttpOnly cookie transport; GraphQL context/resolver integration; Inversify service boundaries; an alternative repository/token-service design; tests for cookie rotation and preventing refresh tokens in JSON.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Do not present the entire upstream invoice app, its original personal project narrative or CI/CD history as Donald’s work. Variants differ; do not combine them into a claim that one final implementation contains all safeguards. Security correctness and concurrency behavior were not independently validated. Skip timesheets, credentials, personal evaluation metadata and real invoice data. Production security is not established.","attribution":"Local variants extend David Andrea/djblackett’s invoice app with bcrypt password checks, signed access tokens, hashed refresh records, rotation/reuse handling, and authentication routes and services.","contribution":"Local variants extend David Andrea/djblackett’s invoice app with bcrypt password checks, signed access tokens, hashed refresh records, rotation/reuse handling, and authentication routes and services.","result":"Authentication implementations span browser cookies, GraphQL context, and persistence, with targeted test source for rotation and keeping refresh tokens out of JSON responses.","name":"Full-stack authentication study","title":"Full-stack authentication study","description":"Alternative native-authentication implementations inside a supplied React/GraphQL invoice application, with JWT access tokens, rotating refresh sessions and cookie-based browser integration.","category":"Software Systems & Automation","technologies":["TypeScript","React","Express","GraphQL","Prisma","PostgreSQL","JWT","Vitest"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["invoice-authentication-catalog-1","invoice-authentication-catalog-2","invoice-authentication-catalog-3","invoice-authentication-catalog-4","invoice-authentication-catalog-5"],"field":"software-systems","rank":5,"rankReason":"Meaningful backend/authentication work, but must be scoped as an adaptation with provisional candidate attribution.","folders":["dataannotation/my-project","dataannotation/my-project copy","dataannotation/invoice-authentication-study/"],"kind":"adaptation","fullStory":false,"limitations":["Do not present the entire upstream invoice app, its original personal project narrative or CI/CD history as Donald’s work.","Variants differ; do not combine them into a claim that one final implementation contains all safeguards.","Security correctness and concurrency behavior were not independently validated.","Skip timesheets, credentials, personal evaluation metadata and real invoice data.","Production security is not established."]},{"id":"cuda-introduction","date":"Project record · Sep 2026","url":"/projects/cuda-introduction","visual":"schema","why":"Introductory CUDA course material for compiling a supplied device-query program and running it through Slurm.","question":"","hard":"CUDA compiler/toolchain, GPU device properties and batch scheduling exposure, as an introductory reference rather than a custom parallel-computing project.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Do not claim custom CUDA kernels, GPU optimization, performance results or original authorship from this folder. Represent as current learning/reference only, if included publicly.","attribution":"An introductory course reference using the supplied CUDA device-query program, compiler setup, and Slurm job files. No custom GPU kernel is claimed.","contribution":"An introductory course reference using the supplied CUDA device-query program, compiler setup, and Slurm job files. No custom GPU kernel is claimed.","result":"A concrete foundation for inspecting GPU device properties and understanding the CUDA build and batch-execution workflow.","name":"CUDA development environment study","title":"CUDA development environment study","description":"Introductory CUDA course material for compiling a supplied device-query program and running it through Slurm.","category":"Software Systems & Automation","technologies":["CUDA","C++","Slurm"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["cuda-introduction-catalog-1","cuda-introduction-catalog-2","cuda-introduction-catalog-3"],"field":"software-systems","rank":6,"rankReason":"Useful learning context but insufficient contribution evidence to rank as a built project.","folders":["uiuc/ECE408"],"kind":"reference","fullStory":false,"limitations":["Do not claim custom CUDA kernels, GPU optimization, performance results or original authorship from this folder.","Represent as current learning/reference only, if included publicly."]},{"id":"blockchain-principles","date":"Project record · Sep 2026","url":"/projects/blockchain-principles","visual":"schema","why":"A local collection of blockchain course slides and notes.","question":"","hard":"Reference material for distributed ledger principles; specific mastered topics or original systems are not established by file possession.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Reference ownership is not original authorship, completion or demonstrated expertise. Files were inventoried, not deeply read or redistributed.","attribution":"A reading reference composed of supplied blockchain course slides and notes, rather than an original implementation.","contribution":"A reading reference composed of supplied blockchain course slides and notes, rather than an original implementation.","result":"A collected set of lecture material on distributed-ledger principles; no authored system or research result is claimed.","name":"Blockchain principles reading collection","title":"Blockchain principles reading collection","description":"A local collection of blockchain course slides and notes.","category":"Software Systems & Automation","technologies":["Distributed systems","Blockchain coursework"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["blockchain-principles-catalog-1","blockchain-principles-catalog-2"],"field":"software-systems","rank":7,"rankReason":"Retain under reading/study material, not among implemented systems.","folders":["uiuc/blockchains_principles"],"kind":"reference","fullStory":false,"limitations":["Reference ownership is not original authorship, completion or demonstrated expertise.","Files were inventoried, not deeply read or redistributed."]},{"id":"kalshibot","date":"September 2026","url":"/projects/kalshibot","visual":"market","why":"Separate execution, experiments, and runtime state so a trading system can be inspected and recovered without treating a plausible backtest as proof.","question":"What happens when an exchange accepts an order but the acknowledgement disappears?","hard":"Partial fills, unknown order intents, restarts, and cancellation ownership all have to preserve quantity-level exposure.","assumptions":"A missing response is an unknown outcome. Passing simulated-client tests does not establish identical live behavior.","data":"Source snapshots, offline fixtures, request/response scenarios, and a separate research protocol. Private orders and account data are not exposed here.","failure":"A clean refactor and a large test suite still cannot prove production equivalence. That acceptance gap remains visible in the project record.","changed":"The large live module was separated by responsibility. Ambiguous order creation is reconciled instead of automatically retried.","lessons":"Separate implementation safety, execution fidelity, and economic validity into different acceptance gates.","unresolved":"Production-input differential replay, account takeover, and exact economic equivalence remain unproven in the inspected evidence.","attribution":"Independent refactor in Donald’s project collection. Source and Git provenance checked; agent assistance and upstream boundaries retained.","name":"Prediction-market execution systems","title":"Prediction-market execution systems","description":"Order reconciliation, execution controls, queue-aware research, and an independently packaged refactor of a prediction-market system.","category":"Quantitative Finance & Markets","technologies":["Python","Reconciliation","Event-driven systems","Offline testing","Queue-aware replay","Qwen","OpenAI-compatible inference"],"roles":["Quant Research","Software Engineering","Research"],"contribution":"Donald’s agent-assisted independent refactor separates execution, tape, backtesting, research, and dashboards, with explicit runtime paths and write controls. Earlier system and research snapshots retain shared-source attribution.","result":"An independently packaged execution/research system with reconciliation for ambiguous order outcomes. Dated September 4 verification records 1,526 passing offline tests; production equivalence and profitability are separate, unproven outcomes.","evidence":["kalshi-refactor","kalshi-ack","kalshi-checks","kalshibot-catalog-1","kalshibot-catalog-2","kalshibot-catalog-3","kalshibot-catalog-4","kalshibot-catalog-5"],"field":"quant-finance","rank":1,"rankReason":"The strongest execution-systems evidence: explicit order lifecycles, failure handling, and independent verification records.","folders":["kalshibot","kalshibot-membrane","kalshibot-membrane-crosssection","kalshibot-membrane-sweepfreeze"],"kind":"built","fullStory":true,"limitations":[]},{"id":"statistical-arbitrage","date":"Project record · Sep 2026","url":"/projects/statistical-arbitrage","visual":"schema","why":"Graph-conditioned residuals, pair selection, and a constrained search harness for testing alternative research modules.","question":"","hard":"Dynamic peer graphs, efficient-price filtering, VECM residual experts, pair-quality admission, expected-net-alpha decomposition, portfolio constraints and a separation between search evaluation and promotion validation.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"The design document mixes plans with implemented work; only source-backed components are claimed complete. Minute-bar search does not establish order-book execution fidelity. No gains from automated search are asserted.","attribution":"Donald implemented graph, residual, expected-net-alpha, pair-selection, and execution layers, plus a bounded task/evaluator pack for exploring alternative research modules.","contribution":"Donald implemented graph, residual, expected-net-alpha, pair-selection, and execution layers, plus a bounded task/evaluator pack for exploring alternative research modules.","result":"Six constrained search tasks sit alongside graph and residual-model components, with separate promotion and full-system validation entrypoints.","name":"Statistical arbitrage & module discovery","title":"Statistical arbitrage & module discovery","description":"Graph-conditioned residuals, pair selection, and a constrained search harness for testing alternative research modules.","category":"Quantitative Finance & Markets","technologies":["Python","Graphs","VECM","Statistical arbitrage","SkyDiscover","Deterministic evaluators"],"roles":["Quant Research","Research"],"evidence":["statistical-arbitrage-catalog-1","statistical-arbitrage-catalog-2","statistical-arbitrage-catalog-3","statistical-arbitrage-catalog-4"],"field":"quant-finance","rank":2,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["pair-trade","sky_discover_pair_trade"],"kind":"built","fullStory":false,"limitations":["The design document mixes plans with implemented work; only source-backed components are claimed complete. Minute-bar search does not establish order-book execution fidelity. No gains from automated search are asserted."]},{"id":"options-variance","date":"Project record · Sep 2026","url":"/projects/options-variance","visual":"schema","why":"Options research and execution tooling spanning implied-versus-realized volatility, short-strangle protocols, and an earnings-event variance decomposition study.","question":"","hard":"Black-Scholes pricing and delta solvers, chain/expiry selection, walk-forward protocols, overlap-aware variance evaluation, tail-risk filters and event-versus-diffusion variance with natural bid/ask fill comparisons.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Not a claim of returns or live deployment. Simplified daily Black-Scholes marks differ from executable option quotes; natural-fill and mid-fill analyses must remain distinct. No broker/account state accessed.","attribution":"Donald implemented pricing, option-chain selection, strategy/backtest modules, and an IBKR API layer. A related earnings-event study separates jump variance from diffusion variance.","contribution":"Donald implemented pricing, option-chain selection, strategy/backtest modules, and an IBKR API layer. A related earnings-event study separates jump variance from diffusion variance.","result":"Reusable options research and execution components support walk-forward protocols and offline tests. The event study specifies prior-only jump estimates, shrinkage, and quote-quality gates.","name":"Options & variance-risk-premium research","title":"Options & variance-risk-premium research","description":"Options research and execution tooling spanning implied-versus-realized volatility, short-strangle protocols, and an earnings-event variance decomposition study.","category":"Quantitative Finance & Markets","technologies":["Python","IBKR API","Options pricing","Variance decomposition","Walk-forward research"],"roles":["Quant Research","Research"],"evidence":["options-variance-catalog-1","options-variance-catalog-2","options-variance-catalog-3","options-variance-catalog-4"],"field":"quant-finance","rank":3,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["IV_trading"],"kind":"built","fullStory":false,"limitations":["Not a claim of returns or live deployment. Simplified daily Black-Scholes marks differ from executable option quotes; natural-fill and mid-fill analyses must remain distinct. No broker/account state accessed."]},{"id":"lppls-research-cockpits","date":"Project record · Sep 2026","url":"/projects/lppls-research-cockpits","visual":"schema","why":"LPPLS bubble detection, an inspectable research cockpit, and related market-mechanism studies with frozen-history checks and rollback controls.","question":"","hard":"Nested trailing-window LPPLS fits; purged out-of-fold meta-models and calibration; PIT entry/label clocks; frozen histories; incremental data refresh with rollback and regression guards; separating timing, stock selection, sizing and hedging; serving prebuilt research artifacts through FastAPI without recomputing the engine on each request.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"This pass inspected documentation and source only; it did not launch or validate current services, scheduled refresh, accounts or model results. Local documents use live for a current cockpit or paper workflow, which must not be presented as verified real-money execution. Historical return/Sortino/capacity figures are not promoted. Cost-basis histograms are vendor-imputed price/turnover proxies, not direct observations of investor intentions.","attribution":"Local implementations connect LPPLS fitting, meta-labeling, validation, backtests, and two research cockpits. The A-share service supports dated research and historical reconciliation; the global monitor fits trailing windows on indices and ETFs.","contribution":"Local implementations connect LPPLS fitting, meta-labeling, validation, backtests, and two research cockpits. The A-share service supports dated research and historical reconciliation; the global monitor fits trailing windows on indices and ETFs.","result":"Inspectable dashboards expose market states, signals, simulated books, performance, and data freshness. Research records preserve rejected mechanisms and checks that stopped unsupported conclusions from advancing.","name":"LPPLS market-state research and interactive cockpits","title":"LPPLS market-state research and interactive cockpits","description":"LPPLS bubble detection, an inspectable research cockpit, and related market-mechanism studies with frozen-history checks and rollback controls.","category":"Quantitative Finance & Markets","technologies":["Python","LPPLS","FastAPI","Purged cross-validation","Isotonic calibration","Parquet","Interactive dashboards"],"roles":["Quant Research","Research"],"evidence":["lppls-research-cockpits-catalog-1","lppls-research-cockpits-catalog-2","lppls-research-cockpits-catalog-3","lppls-research-cockpits-catalog-4","lppls-research-cockpits-catalog-5","lppls-research-cockpits-catalog-6","lppls-research-cockpits-catalog-7","lppls-research-cockpits-catalog-8","lppls-research-cockpits-catalog-9"],"field":"quant-finance","rank":4,"rankReason":"A strong research-system bridge: a detector, inspectable dashboards, regression safeguards, and documented falsification.","folders":["A_Share/openassetpricing/metalabel","A_Share/openassetpricing/global_bubble","A_Share/openassetpricing/frame_attack","A_Share/openassetpricing/tail_runway","A_Share/openassetpricing/pm_system","A_Share/openassetpricing/report/cbsf_factor"],"kind":"built","fullStory":false,"limitations":["This pass inspected documentation and source only; it did not launch or validate current services, scheduled refresh, accounts or model results. Local documents use live for a current cockpit or paper workflow, which must not be presented as verified real-money execution. Historical return/Sortino/capacity figures are not promoted. Cost-basis histograms are vendor-imputed price/turnover proxies, not direct observations of investor intentions."]},{"id":"zero-human-hedge","date":"Project record · Sep 2026","url":"/projects/zero-human-hedge","visual":"schema","why":"An integrated paper-research stack connecting a minute-data lab, portfolio optimizer, Nautilus parity backtest, observability and an agent control room.","question":"","hard":"Research artifact manifests, session normalization, feature sweeps, constrained targets, backtest parity, DuckDB reporting, Prometheus/Grafana metrics and Paperclip integration.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Paper fund is an experiment/operations metaphor, not evidence of managing a fund or client money. No autonomous profitability, deployed service status or external user claim.","attribution":"Donald connected data aggregation, feature research, portfolio optimization, parity backtesting, and report synchronization through CLI and application integrations.","contribution":"Donald connected data aggregation, feature research, portfolio optimization, parity backtesting, and report synchronization through CLI and application integrations.","result":"A paper-research operations stack with artifact manifests, Docker infrastructure, dashboards, and observability spanning the research loop.","name":"Zero Human Hedge — paper research operations","title":"Zero Human Hedge — paper research operations","description":"An integrated paper-research stack connecting a minute-data lab, portfolio optimizer, Nautilus parity backtest, observability and an agent control room.","category":"Quantitative Finance & Markets","technologies":["Python","DuckDB","NautilusTrader","Paperclip","Docker","Prometheus","Grafana"],"roles":["Quant Research","Research"],"evidence":["zero-human-hedge-catalog-1","zero-human-hedge-catalog-2","zero-human-hedge-catalog-3"],"field":"quant-finance","rank":5,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["zerohumanhedge"],"kind":"built","fullStory":false,"limitations":["Paper fund is an experiment/operations metaphor, not evidence of managing a fund or client money. No autonomous profitability, deployed service status or external user claim."]},{"id":"mention-market-model","date":"Project record · Sep 2026","url":"/projects/mention-market-model","visual":"schema","why":"A distinct event-market study combining broadcaster transcripts, settlement labels, crew effects and mention-time approximations.","question":"","hard":"Beta-smoothed word baselines, hierarchical crew/team effects, empirical-Bayes shrinkage, knockout-regime changes, bounded corpus-logit tilts and leave-one-out evaluation with corpus-leakage adjustments.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Treat this as a dated July 2026 study. Captions and highlights have unequal coverage; first-mention timing is an approximation from price paths and excludes unsuitable contexts. No trading edge or out-of-sample deployment claim.","attribution":"A standalone local study implements transcript-corpus ingestion, settlement-label processing, mention-time approximations, and hierarchical broadcaster/team effects, separate from the upstream Kalshi bot.","contribution":"A standalone local study implements transcript-corpus ingestion, settlement-label processing, mention-time approximations, and hierarchical broadcaster/team effects, separate from the upstream Kalshi bot.","result":"The documented study found crew-settlement effects more informative than corpus tilts for words with settlement history. The result is a dated research finding, not an established trading edge.","name":"World Cup announcer-mention modeling","title":"World Cup announcer-mention modeling","description":"A distinct event-market study combining broadcaster transcripts, settlement labels, crew effects and mention-time approximations.","category":"Quantitative Finance & Markets","technologies":["Python","NLP corpora","Empirical Bayes","Beta-binomial models","Calibration"],"roles":["Quant Research","Research"],"evidence":["mention-market-model-catalog-1","mention-market-model-catalog-2","mention-market-model-catalog-3"],"field":"quant-finance","rank":6,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["prediction_market"],"kind":"study","fullStory":false,"limitations":["Treat this as a dated July 2026 study. Captions and highlights have unequal coverage; first-mention timing is an approximation from price paths and excludes unsuitable contexts. No trading edge or out-of-sample deployment claim."]},{"id":"factor-replication","date":"Project record · Sep 2026","url":"/projects/factor-replication","visual":"schema","why":"Asset-pricing factor replication, index-enhancement experiments, and related upstream data infrastructure.","question":"","hard":"Announcement-date availability, monthly universes, market-specific tradability, factor preprocessing, cross-sectional neutralization, factor aggregation, portfolio formation and robust inference.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"212 is the OpenAP library scope, not a verified successful-factor count. No profitable factor discovery, exact reproduction of every definition or Hikyuu authorship/release/adoption claim.","attribution":"Local factor implementations and evaluation harnesses adapt published predictor definitions to A-share availability and trading constraints. Hikyuu remains upstream infrastructure with a scoped local import change.","contribution":"Local factor implementations and evaluation harnesses adapt published predictor definitions to A-share availability and trading constraints. Hikyuu remains upstream infrastructure with a scoped local import change.","result":"The OpenAP harness covers a documented library of 212 predictor definitions, alongside ingestion, factor evaluation, and portfolio-formation workflows. Library scope is not a successful-factor count.","name":"A-share factor replication and evaluation","title":"A-share factor replication and evaluation","description":"Asset-pricing factor replication, index-enhancement experiments, and related upstream data infrastructure.","category":"Quantitative Finance & Markets","technologies":["Python","Tushare","Asset pricing","Factor models","PIT universes","Statistical inference","C++","HDF5","Market-data ingestion"],"roles":["Quant Research","Research"],"evidence":["factor-replication-catalog-1","factor-replication-catalog-2","factor-replication-catalog-3","factor-replication-catalog-4","factor-replication-catalog-5"],"field":"quant-finance","rank":7,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/openassetpricing","A_Share/factorResearch/paper1","A_Share/factorResearch-research","A_Share/hikyuu"],"kind":"adaptation","fullStory":false,"limitations":["212 is the OpenAP library scope, not a verified successful-factor count. No profitable factor discovery, exact reproduction of every definition or Hikyuu authorship/release/adoption claim."]},{"id":"mean-reversion-system","date":"Project record · Sep 2026","url":"/projects/mean-reversion-system","visual":"schema","why":"A short-horizon reversal system with clustering, market-specific tradeability, portfolio construction and futures hedging.","question":"","hard":"Lagged signals and next-open execution, overlapping holding periods, clustering-augmented reversals, continuous regime scaling, partial fills, futures beta hedges and factor-alpha diagnostics.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Research benchmark numbers in documentation are comparison targets, not achieved project returns. Exact historical execution and paper replication were not rerun; the design filename has an inconsistent year and should not determine public chronology.","attribution":"Donald implemented alpha, universe, feature, portfolio, execution, and backtest modules, including clustering-enhanced reversal signals and market-specific tradeability checks.","contribution":"Donald implemented alpha, universe, feature, portfolio, execution, and backtest modules, including clustering-enhanced reversal signals and market-specific tradeability checks.","result":"The research system connects lagged signals to next-open execution, overlapping holdings, futures hedges, and factor-alpha diagnostics, with execution-simulator test coverage in source.","name":"A-share reversal with executable constraints","title":"A-share reversal with executable constraints","description":"A short-horizon reversal system with clustering, market-specific tradeability, portfolio construction and futures hedging.","category":"Quantitative Finance & Markets","technologies":["Python","Clustering","Mean reversion","Portfolio optimization","Execution simulation"],"roles":["Quant Research","Research"],"evidence":["mean-reversion-system-catalog-1","mean-reversion-system-catalog-2","mean-reversion-system-catalog-3"],"field":"quant-finance","rank":8,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["mean-reversion"],"kind":"built","fullStory":false,"limitations":["Research benchmark numbers in documentation are comparison targets, not achieved project returns. Exact historical execution and paper replication were not rerun; the design filename has an inconsistent year and should not determine public chronology."]},{"id":"market-signal-ingestion","date":"Project record · Sep 2026","url":"/projects/market-signal-ingestion","visual":"schema","why":"A multi-source signal-ingestion, replay and alerting layer for prediction-market research.","question":"","hard":"REST/WebSocket adapters, normalized source ingestion, reconnect/authentication handling, stream capture, replay and downstream alert interfaces.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"A downstream trading engine is explicitly outside this repository. No signal quality, commercial access entitlement or trading-performance claim. Do not expose session tokens or account integration details publicly.","attribution":"Donald built source adapters, stream capture and replay, downstream alert sinks, and a live-source dashboard.","contribution":"Donald built source adapters, stream capture and replay, downstream alert sinks, and a live-source dashboard.","result":"Normalized REST/WebSocket ingestion covers prediction markets, trading venues, and public disclosures, with reconnect handling and replayable streams for downstream research.","name":"Market-signal ingestion and replay","title":"Market-signal ingestion and replay","description":"A multi-source signal-ingestion, replay and alerting layer for prediction-market research.","category":"Quantitative Finance & Markets","technologies":["Python","httpx","WebSockets","REST APIs","Replay","Dashboards"],"roles":["Quant Research","Research"],"evidence":["market-signal-ingestion-catalog-1","market-signal-ingestion-catalog-2","market-signal-ingestion-catalog-3"],"field":"quant-finance","rank":9,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["poly"],"kind":"built","fullStory":false,"limitations":["A downstream trading engine is explicitly outside this repository. No signal quality, commercial access entitlement or trading-performance claim. Do not expose session tokens or account integration details publicly."]},{"id":"technical-indicator-evaluation","date":"Project record · Sep 2026","url":"/projects/technical-indicator-evaluation","visual":"schema","why":"A repeatable multi-fold evaluation and dashboard system for technical signals on US minute bars.","question":"","hard":"Fold coverage planning, deterministic sample selection, indicator isolation, timeouts, failure replay, rank-information coefficients, cross-fold aggregation and checkpointed execution.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"","attribution":"Local implementations define a common indicator contract and connect discovery, isolated evaluation, failure replay, checkpointing, and serving.","contribution":"Local implementations define a common indicator contract and connect discovery, isolated evaluation, failure replay, checkpointing, and serving.","result":"An evaluator, API, and Streamlit workbench organize a documented library of 100 TradingView-derived indicators across repeatable folds and aggregated metrics.","name":"Technical-indicator evaluation workbench","title":"Technical-indicator evaluation workbench","description":"A repeatable multi-fold evaluation and dashboard system for technical signals on US minute bars.","category":"Quantitative Finance & Markets","technologies":["Python","pandas","NumPy","Streamlit","Parquet","Factor evaluation"],"roles":["Quant Research","Research"],"evidence":["technical-indicator-evaluation-catalog-1","technical-indicator-evaluation-catalog-2"],"field":"quant-finance","rank":10,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["indicator_test"],"kind":"built","fullStory":false,"limitations":[]},{"id":"market-mechanism-studies","date":"Project record · Sep 2026","url":"/projects/market-mechanism-studies","visual":"schema","why":"Studies of market bottoms, public flow disclosures, IPOs, and observable policy-support signals, with rejected hypotheses retained.","question":"","hard":"Availability clocks, event and era definitions, frozen panels, feasible-trade constraints, robust ETF-flow proxies, missing-feed confidence, and converting informal mechanisms into measurable tests.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Report findings are dated and not independently reproduced. No financial figures promoted. Policy scores identify observable patterns, not actual buyer identity or intent. IPO listing returns do not establish achievable subscription returns.","attribution":"Local experiment harnesses, reports, and a Streamlit scoring engine turn event, disclosure, and policy-support hypotheses into measurable tests.","contribution":"Local experiment harnesses, reports, and a Streamlit scoring engine turn event, disclosure, and policy-support hypotheses into measurable tests.","result":"The studies preserve rejected hypotheses and test the limits of disclosure following. A seven-component tracker scores observable policy-support patterns, with its synthetic demonstration labeled explicitly.","name":"A-share event, flow and intervention studies","title":"A-share event, flow and intervention studies","description":"Studies of market bottoms, public flow disclosures, IPOs, and observable policy-support signals, with rejected hypotheses retained.","category":"Quantitative Finance & Markets","technologies":["Python","Event studies","Tushare","Time-series analysis","Hypothesis testing","Streamlit","Robust statistics","Interpretable scoring"],"roles":["Quant Research","Research"],"evidence":["market-mechanism-studies-catalog-1","market-mechanism-studies-catalog-2","market-mechanism-studies-catalog-3","market-mechanism-studies-catalog-4","market-mechanism-studies-catalog-5","market-mechanism-studies-catalog-6"],"field":"quant-finance","rank":11,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["A_Share/SW_research","A_Share/hm_study","A_Share/ipo_study","alpha_testing","china_national_team_tracker"],"kind":"study","fullStory":false,"limitations":["Report findings are dated and not independently reproduced. No financial figures promoted. Policy scores identify observable patterns, not actual buyer identity or intent. IPO listing returns do not establish achievable subscription returns."]},{"id":"finance-reading-and-falsification","date":"Project record · Sep 2026","url":"/projects/finance-reading-and-falsification","visual":"schema","why":"Research notes that turn trading literature and informal claims into testable rules, experiment plans, and critical assessments.","question":"","hard":"Separating descriptive patterns from predictive claims, expressing rule availability, identifying tests that require unavailable data, outlining ablations and judging falsifiability.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Literature performance claims are external claims, not Donald’s results. Reading-material presence does not establish personal beliefs or endorsements. No full books or copied teaching material should be republished.","attribution":"Research artifacts translate trading literature into a long-only blueprint, testable mechanical claims, experiment plans, and critical assessments. They are not presented as Donald’s verified first-person writing.","contribution":"Research artifacts translate trading literature into a long-only blueprint, testable mechanical claims, experiment plans, and critical assessments. They are not presented as Donald’s verified first-person writing.","result":"The notebooks separate descriptive patterns from predictive claims and identify which tests require unavailable evidence. The Al Brooks study produces a report and blueprint rather than an implemented strategy.","name":"Trading literature and falsification notebooks","title":"Trading literature and falsification notebooks","description":"Research notes that turn trading literature and informal claims into testable rules, experiment plans, and critical assessments.","category":"Quantitative Finance & Markets","technologies":["Literature synthesis","Experiment design","Falsification","Quantitative research"],"roles":["Quant Research","Research"],"evidence":["finance-reading-and-falsification-catalog-1","finance-reading-and-falsification-catalog-2","finance-reading-and-falsification-catalog-3","finance-reading-and-falsification-catalog-4"],"field":"quant-finance","rank":12,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["strategy_momentum","A_Share/al_brooks","A_Share/yi","A_Share/insider_case_study"],"kind":"study","fullStory":false,"limitations":["Literature performance claims are external claims, not Donald’s results. Reading-material presence does not establish personal beliefs or endorsements. No full books or copied teaching material should be republished."]},{"id":"csc-tactics","date":"September 2026","url":"/projects/csc-tactics","visual":"pitch","why":"Make a tactical situation understandable from different teammates’ viewpoints, then connect that explanation to practice.","question":"Can the same tactical idea make sense from eleven positions on the field?","hard":"Changing a drill changes scene objects, ball counts, and camera state together. Both renderers need consistent behavior across transitions.","assumptions":"Screen-based learning is a teaching aid. It is not evidence that a score transfers to match performance.","data":"52 tactical challenges, 11 roles, eight drills, and 43 training segments. Chinese training content is preserved in the original application.","failure":"A two-ball drill changed to a one-ball drill while a stale extra-ball mesh remained. A camera reset rendered before the scene was synchronized.","changed":"Scene meshes are synchronized before the camera reset can trigger a render.","lessons":"A scene is not finished when the first frame looks right. Test the transitions between scenes.","unresolved":"How well screen explanations transfer to decisions on an actual field is not measured.","attribution":"Donald’s project collection; project content and renderer documentation inspected. No invented coaching role or team biography.","name":"CSC:tactics","title":"CSC:tactics","description":"A Chinese soccer tactics teaching application with 52 situational challenges, 11 roles, three viewpoints and a 90-minute training plan containing 43 segments.","category":"Interactive Tools, Learning & Writing","technologies":["JavaScript","Three.js","Canvas","Vite","Interactive learning"],"roles":["Software Engineering","Something else"],"contribution":"The project combines Chinese teaching content, 52 situational challenges, 11 selectable roles, three viewpoints, and a structured training plan using shared Three.js/Canvas scene logic.","result":"A complete teaching application connects tactical explanations to a 90-minute practice plan with 43 segments. September 2026 verification records 11 Node and 12 browser tests, including both renderers and every training segment.","evidence":["soccer-content","soccer-lifecycle","soccer-reflection","csc-tactics-catalog-1","csc-tactics-catalog-2","csc-tactics-catalog-3","csc-tactics-catalog-4"],"field":"creative-tools","rank":2,"rankReason":"The most complete interactive teaching tool: tactical challenges, animated drills, role perspectives, and a rendering fallback.","folders":["soccer"],"kind":"built","fullStory":true,"limitations":["Screen learning is not validated improvement in match performance.","No player/user counts or coaching biography established."]},{"id":"youtube-knowledge","date":"Project record · Sep 2026","url":"/projects/youtube-knowledge","visual":"schema","why":"A Chinese-first reading and exploration application that turns video transcripts into staged summaries, source-linked ideas, and creator-level knowledge graphs.","question":"","hard":"Fast summary and deeper analysis are separated. A bounded semaphore permits three ingests, per-channel locks protect shared profiles, and temp-file replacement keeps writes atomic. NDJSON heartbeats keep long jobs observable. Graph validation restricts source video IDs and requires an actual cross-video connection; graph failure does not discard primary distillation.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Transcript content and creator conclusions are external material, not Donald-authored beliefs or writing. Uses Claude CLI for analysis; this is not local Qwen inference. Live uptime, extraction accuracy, user adoption, and all historical browser outcomes were not reverified. Only implementation and test source were inspected; stored transcripts, chat logs, and personal library contents were excluded.","attribution":"Donald built transcript ingestion, streaming APIs, source-linked distillation, creator knowledge graphs, and query/feed views, then improved concurrency and mobile behavior.","contribution":"Donald built transcript ingestion, streaming APIs, source-linked distillation, creator knowledge graphs, and query/feed views, then improved concurrency and mobile behavior.","result":"A Chinese-first application connects videos to summaries and cross-video ideas. Bounded ingestion, per-channel locking, atomic writes, and validated source IDs keep long-running knowledge workflows observable and recoverable.","name":"YouTube Knowledge","title":"YouTube Knowledge","description":"A Chinese-first reading and exploration application that turns video transcripts into staged summaries, source-linked ideas, and creator-level knowledge graphs.","category":"Interactive Tools, Learning & Writing","technologies":["Python","Claude CLI","NDJSON","Threads","Atomic file replacement","JSON","YouTube transcripts","JavaScript"],"roles":["Software Engineering"],"evidence":["youtube-knowledge-catalog-1","youtube-knowledge-catalog-2","youtube-knowledge-catalog-3","youtube-knowledge-catalog-4","youtube-knowledge-catalog-5","youtube-knowledge-catalog-6"],"field":"creative-tools","rank":3,"rankReason":"Strong applied product: substantial Donald history, streaming/concurrency details, source-linked information architecture, and actual workflow use.","folders":["youtube_knowledge"],"kind":"built","fullStory":false,"limitations":["Transcript content and creator conclusions are external material, not Donald-authored beliefs or writing.","Uses Claude CLI for analysis; this is not local Qwen inference.","Live uptime, extraction accuracy, user adoption, and all historical browser outcomes were not reverified.","Only implementation and test source were inspected; stored transcripts, chat logs, and personal library contents were excluded."]},{"id":"piano-midi","date":"February 2026 onward","url":"/projects/piano-midi","visual":"midi","why":"Turn a YouTube or Bilibili piano video into a downloadable MIDI file and an organized conversion library.","question":"How much useful cleanup belongs after transcription?","hard":"Downloaded audio, model failures, GPU availability, overlapping notes, and variable timing need a coherent conversion flow.","assumptions":"Audio transcription is provided by Transkun. The project does not claim a new transcription model.","data":"User-supplied video/audio, predicted MIDI, estimated tempo, timing and velocity values. No private library or copyrighted recording is republished.","failure":"CUDA transcription can fail; the implementation retries on CPU. Musical transcription accuracy has not been independently evaluated.","changed":"Postprocessing adds partial sixteenth-note quantization, velocity smoothing, and same-pitch overlap trimming.","lessons":"The useful product includes the handling around the model: input, recovery, cleanup, and retrieval.","unresolved":"Transcription accuracy across recording conditions and the tradeoff between timing cleanup and expressive performance need evaluation.","attribution":"Integration and product engineering; Transkun is an upstream dependency. The project is not evidence of Donald’s musical biography.","name":"Piano video to MIDI","title":"Piano video to MIDI","description":"A web pipeline converting YouTube/Bilibili piano audio into editable MIDI, with optional timing/dynamics cleanup and a saved library.","category":"Interactive Tools, Learning & Writing","technologies":["Python","Flask","Transkun","MIDI","PyTorch","pretty-midi","yt-dlp"],"roles":["Software Engineering","AI / ML","Something else"],"contribution":"Donald built video ingestion, job orchestration, library deduplication, and MIDI postprocessing around upstream Transkun transcription.","result":"A conversion workflow with serial queueing, CPU fallback after CUDA failure, optional timing and velocity cleanup, overlap trimming, and a saved MIDI library.","evidence":["piano-pipeline","piano-midi-catalog-1","piano-midi-catalog-2","piano-midi-catalog-3","piano-midi-catalog-4"],"field":"creative-tools","rank":4,"rankReason":"Distinct creative integration with good candidate attribution and an understandable end-to-end function.","folders":["pianovision_midi_converter"],"kind":"built","fullStory":true,"limitations":["Transkun is upstream, not a model Donald trained.","No hand-splitting implementation found.","The project does not establish that Donald plays piano or owns PianoVision.","User library/audio files were not inspected.","Transcription accuracy remains unmeasured."]},{"id":"browser-repl","date":"Project record · Sep 2026","url":"/projects/browser-repl","visual":"schema","why":"Two related Vue/CodeMirror implementations of a browser-only JavaScript/Python editor with execution, console output, saved files and export.","question":"","hard":"CodeMirror language extensions and completions; browser execution bridges and console serialization; lazy Pyodide/WebAssembly loading; worker lifecycle and execution timeouts; localStorage file management; formatting and export; draggable output layout.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Origin is an evaluation/project folder; avoid implying paid employment or customer delivery. Do not call either variant a production-secure sandbox. The iframe/main-thread version cannot reliably interrupt blocking execution; worker version has explicit termination budgets. No authored commit history or usability/compatibility measurements verified.","attribution":"Local Vue/CodeMirror variants implement editing, language switching, file persistence, console output, and export. One uses an iframe/main-thread runtime; the other runs JavaScript and Python in terminable workers.","contribution":"Local Vue/CodeMirror variants implement editing, language switching, file persistence, console output, and export. One uses an iframe/main-thread runtime; the other runs JavaScript and Python in terminable workers.","result":"Two browser-only execution designs explore the same editor workflow, including lazy Pyodide loading and worker execution budgets. They are experiments rather than production-secure sandboxes.","name":"Sandbox: browser code experiments","title":"Sandbox: browser code experiments","description":"Two related Vue/CodeMirror implementations of a browser-only JavaScript/Python editor with execution, console output, saved files and export.","category":"Interactive Tools, Learning & Writing","technologies":["Vue","CodeMirror","Vite","Web Workers","Pyodide","WebAssembly"],"roles":["Software Engineering"],"evidence":["browser-repl-catalog-1","browser-repl-catalog-2","browser-repl-catalog-3","browser-repl-catalog-4"],"field":"creative-tools","rank":5,"rankReason":"Meaningful application and runtime tradeoffs; should be visible as an experiment family, with weaker authorship evidence than the two preceding projects.","folders":["dataannotation/may22","dataannotation/may22codex"],"kind":"built","fullStory":false,"limitations":["Origin is an evaluation/project folder; avoid implying paid employment or customer delivery.","Do not call either variant a production-secure sandbox.","The iframe/main-thread version cannot reliably interrupt blocking execution; worker version has explicit termination budgets.","No authored commit history or usability/compatibility measurements verified."]},{"id":"daily-news-terminal","date":"Project record · Sep 2026","url":"/projects/daily-news-terminal","visual":"schema","why":"A dockerized live-information interface combining news signals, prediction-market context, public trade activity, and ADS-B flight overlays.","question":"","hard":"A Node server normalizes heterogeneous public sources, labels source warnings and stale cache, scores textual relevance to markets, maintains keyword alerts, and pushes WebSocket updates to a globe and dense feed interface.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Candidate-specific authorship is not corroborated by version history. Market matching is an entity/token heuristic, not demonstrated LLM reasoning or causal inference. No accuracy, reliability, live uptime, commercial adoption, or trading performance was established. Optional alerts and configured external integrations were not invoked.","attribution":"The local implementation normalizes public feeds, scores text-to-market relevance, manages keyword alerts, and streams updates to a globe and dense news interface.","contribution":"The local implementation normalizes public feeds, scores text-to-market relevance, manages keyword alerts, and streams updates to a globe and dense news interface.","result":"A Docker-packaged server and frontend combine news, prediction-market context, trade activity, and flight overlays. Source warnings and stale-cache labels replace built-in fake-data fallbacks.","name":"Daily News Vision Terminal","title":"Daily News Vision Terminal","description":"A dockerized live-information interface combining news signals, prediction-market context, public trade activity, and ADS-B flight overlays.","category":"Interactive Tools, Learning & Writing","technologies":["Node.js","JavaScript","WebSocket","Docker","GDELT","Polymarket public APIs","OpenSky ADS-B"],"roles":["Software Engineering"],"evidence":["daily-news-terminal-catalog-1","daily-news-terminal-catalog-2","daily-news-terminal-catalog-3","daily-news-terminal-catalog-4"],"field":"creative-tools","rank":6,"rankReason":"Keep breadth visible as an inspectable live-data UI/system; no Git provenance means authorship should be confirmed before headline ownership claims.","folders":["daily-news-terminal-docker"],"kind":"built","fullStory":false,"limitations":["Candidate-specific authorship is not corroborated by version history.","Market matching is an entity/token heuristic, not demonstrated LLM reasoning or causal inference.","No accuracy, reliability, live uptime, commercial adoption, or trading performance was established.","Optional alerts and configured external integrations were not invoked."]},{"id":"video-transcriber-adaptation","date":"Project record · Sep 2026","url":"/projects/video-transcriber-adaptation","visual":"schema","why":"Local adaptations to an upstream subtitle-first transcription and summarization application, including a different model client, GPU transcription and platform-specific retrieval handling.","question":"","hard":"Async model-client compatibility layer collecting streamed text; model configuration normalization; subtitle-first versus audio fallback pipeline; GPU speech recognition; platform extraction and smoke scripts.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Upstream app authorship remains with its original contributors. Model-list strings are implementation configuration, not independent confirmation of provider product availability. No live platform extraction, model call or GPU benchmark was run. Transcript/media temp files and credentials were not inspected. Throughput and transcription accuracy have not been benchmarked.","attribution":"Local changes extend wendy7756’s AI-Video-Transcriber with a Claude Agent SDK adapter, shared model configuration, CUDA float16 Faster-Whisper, and Bilibili retrieval handling.","contribution":"Local changes extend wendy7756’s AI-Video-Transcriber with a Claude Agent SDK adapter, shared model configuration, CUDA float16 Faster-Whisper, and Bilibili retrieval handling.","result":"The adaptation connects streamed model output and GPU speech recognition to an upstream subtitle-first workflow.","name":"Video transcription workflow adaptation","title":"Video transcription workflow adaptation","description":"Local adaptations to an upstream subtitle-first transcription and summarization application, including a different model client, GPU transcription and platform-specific retrieval handling.","category":"Interactive Tools, Learning & Writing","technologies":["Python","FastAPI","Faster-Whisper","CUDA","yt-dlp","Claude Agent SDK"],"roles":["Software Engineering"],"evidence":["video-transcriber-adaptation-catalog-1","video-transcriber-adaptation-catalog-2","video-transcriber-adaptation-catalog-3","video-transcriber-adaptation-catalog-4","video-transcriber-adaptation-catalog-5"],"field":"creative-tools","rank":7,"rankReason":"Real integration work, appropriately secondary because the app is an upstream clone and local contribution attribution is weaker.","folders":["video_transribe/AI-Video-Transcriber"],"kind":"adaptation","fullStory":false,"limitations":["Upstream app authorship remains with its original contributors.","Model-list strings are implementation configuration, not independent confirmation of provider product availability.","No live platform extraction, model call or GPU benchmark was run.","Transcript/media temp files and credentials were not inspected.","Throughput and transcription accuracy have not been benchmarked."]},{"id":"personal-website","date":"Project record · Sep 2026","url":"/projects/personal-website","visual":"schema","why":"This site: a field-based project catalog and a local-Qwen guide that answers from reviewed claims and linked evidence.","question":"","hard":"A common knowledge model for pages and machine-readable feeds; schema-constrained claim selection; server-side inference; signed answer snapshots; persistent sharing; reduced-motion spatial interaction.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"AI-assisted implementation. The guide selects reviewed claims rather than reproducing Donald’s voice; missing biographical information remains unknown.","attribution":"Donald directed the information model and design, including the discipline-based reorganization. The implementation was built with an AI coding agent: shared structured content, ranked project families, a local-Qwen evidence guide and immutable shared answers.","contribution":"Donald directed the information model and design, including the discipline-based reorganization. The implementation was built with an AI coding agent: shared structured content, ranked project families, a local-Qwen evidence guide and immutable shared answers.","result":"An implemented website with field navigation, searchable projects, detailed evidence, recruiter views, JSON feeds and a local-Qwen answer service.","name":"A searchable engineering notebook","title":"A searchable engineering notebook","description":"This site: a field-based project catalog and a local-Qwen guide that answers from reviewed claims and linked evidence.","category":"Interactive Tools, Learning & Writing","technologies":["React","TypeScript","Cloudflare","Local Qwen","Semantic HTML"],"roles":["Software Engineering"],"evidence":["personal-website-catalog-1","personal-website-catalog-2","personal-website-catalog-3"],"field":"creative-tools","rank":8,"rankReason":"Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.","folders":["personal-website"],"kind":"built","fullStory":false,"limitations":["AI-assisted implementation. The guide selects reviewed claims rather than reproducing Donald’s voice; missing biographical information remains unknown."]},{"id":"technology-society-writing","date":"Project record · Sep 2026","url":"/projects/technology-society-writing","visual":"schema","why":"A candidate-attributed course essay examines American Graffiti through histories of mobility, vehicle identity and the social consequences of transport technology.","question":"","hard":"Evidence-based synthesis across a film and historical readings; distinguishing the promises of technological progress from its social effects; an academic writing artifact that broadens the engineering picture.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Do not infer favorite films, political views, personal motivations or identity traits from the essay. Do not republish long excerpts from quoted books or the film. Fordlandia exists as a PDF but its authorship/content was not inspected; list as a related local artifact only.","attribution":"Donald’s course essay connects American Graffiti to historical readings on American mobility, the Yugo, decorated trucks, and railroads.","contribution":"Donald’s course essay connects American Graffiti to historical readings on American mobility, the Yugo, decorated trucks, and railroads.","result":"An authored essay examining how transport technology shapes social life and identity, broadening the collection beyond engineering implementation.","name":"Cars, identity & technological change","title":"Cars, identity & technological change","description":"A candidate-attributed course essay examines American Graffiti through histories of mobility, vehicle identity and the social consequences of transport technology.","category":"Interactive Tools, Learning & Writing","technologies":["Academic writing","Technology history","LaTeX"],"roles":["Software Engineering"],"evidence":["technology-society-writing-catalog-1","technology-society-writing-catalog-2"],"field":"creative-tools","rank":9,"rankReason":"Strongest newly found nontechnical breadth evidence; use a small writing/curiosity entry, not a fabricated personality claim.","folders":["uiuc/hist"],"kind":"study","fullStory":false,"limitations":["Do not infer favorite films, political views, personal motivations or identity traits from the essay.","Do not republish long excerpts from quoted books or the film.","Fordlandia exists as a PDF but its authorship/content was not inspected; list as a related local artifact only."]},{"id":"illinix-391","date":"Spring 2025","url":"/projects/illinix-391","location":"Champaign, IL","visual":"schema","why":"A Unix-like kernel in C and RISC-V assembly, from virtual memory to an interactive shell.","question":"","hard":"ELF binary loading, preemptive multitasking, UNIX pipes, and shell I/O redirection connect process execution, scheduling, storage, and device I/O into one operating system.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Project architecture and scope are from Donald’s Spring 2025 project summary.","attribution":"Architected a kernel from scratch with virtual memory, fork/exec, 15+ system calls, VIRTIO block drivers, and a custom KTFS filesystem.","contribution":"Architected a kernel from scratch with virtual memory, fork/exec, 15+ system calls, VIRTIO block drivers, and a custom KTFS filesystem.","result":"A complete kernel supporting process creation, executable loading, filesystem access, multitasking, and an interactive shell.","name":"Illinix 391","title":"Illinix 391","description":"A Unix-like kernel in C and RISC-V assembly, from virtual memory to an interactive shell.","category":"Software Systems & Automation","technologies":["C","RISC-V Assembly","Virtual memory","Operating systems","VIRTIO","KTFS"],"roles":["Software Engineering","Embedded / Vision"],"evidence":["illinix-391-catalog-1"],"field":"software-systems","rank":1,"rankReason":"A complete operating-system stack: memory, processes, devices, storage, and a shell, implemented together.","folders":[],"kind":"built","fullStory":false,"limitations":["Project architecture and scope are from Donald’s Spring 2025 project summary."]},{"id":"reaction-wheel-pendulum","date":"Fall 2024","url":"/projects/reaction-wheel-pendulum","location":"Champaign, IL","visual":"schema","why":"Real-time stabilization of an inherently unstable pendulum using a reaction wheel.","question":"","hard":"The controller combines a nonlinear mechanical model, state feedback, observer-based state estimation, and real-time actuation through Wincon.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Control outcome and implementation details are from Donald’s Fall 2024 project summary.","attribution":"Modeled the system using Lagrangian dynamics and implemented three-state feedback with a Luenberger observer and friction compensation.","contribution":"Modeled the system using Lagrangian dynamics and implemented three-state feedback with a Luenberger observer and friction compensation.","result":"Achieved stable equilibrium control of the unstable nonlinear system.","name":"Reaction Wheel Pendulum","title":"Reaction Wheel Pendulum","description":"Real-time stabilization of an inherently unstable pendulum using a reaction wheel.","category":"Robotics & Autonomous Driving","technologies":["Control systems","Lagrangian dynamics","State feedback","Luenberger observer","Wincon"],"roles":["Robotics / Autonomy","Research"],"evidence":["reaction-wheel-pendulum-catalog-1"],"field":"robotics","rank":2,"rankReason":"Connects nonlinear dynamics, state estimation, and feedback to real-time equilibrium control.","folders":[],"kind":"built","fullStory":false,"limitations":["Control outcome and implementation details are from Donald’s Fall 2024 project summary."]},{"id":"twentyfourpoints","date":"Summer 2025","url":"/projects/twentyfourpoints","location":"Irving, CA","visual":"schema","why":"A real-time multiplayer math arena, built and launched in 72 hours.","question":"","hard":"Mathematical validation algorithms and auto-balancing support competitive play. The delivery includes unit tests, migration scripts, and cross-device deployment.","assumptions":"","data":"","failure":"","changed":"","lessons":"","unresolved":"Launch time, puzzle count, and uptime are candidate-reported Summer 2025 outcomes. The uptime observation interval was not specified; it is not a current service guarantee. Puzzle counts are not user counts.","attribution":"Built the React/TypeScript client, Node.js and Socket.io multiplayer backend, Supabase authentication and ELO rankings, and a custom mathematical game engine.","contribution":"Built the React/TypeScript client, Node.js and Socket.io multiplayer backend, Supabase authentication and ELO rankings, and a custom mathematical game engine.","result":"Launched in 72 hours. Donald’s Summer 2025 project record reports 760+ unique puzzles solved and 100% uptime during the reported launch period.","name":"twentyfourpoints.com","title":"twentyfourpoints.com","description":"A real-time multiplayer math arena, built and launched in 72 hours.","category":"Interactive Tools, Learning & Writing","technologies":["React","TypeScript","Node.js","Socket.io","Supabase","ELO"],"roles":["Software Engineering"],"evidence":["twentyfourpoints-catalog-1"],"field":"creative-tools","rank":1,"rankReason":"A complete multiplayer product, from mathematical validation to live sessions, authentication, and rankings.","folders":[],"kind":"built","fullStory":false,"limitations":["Launch time, puzzle count, and uptime are candidate-reported Summer 2025 outcomes. The uptime observation interval was not specified; it is not a current service guarantee. Puzzle counts are not user counts."]}],"claims":[{"id":"identity","kind":"FACT","text":"Donald Shen studies Computer Engineering at UIUC. He can be reached at donald7@illinois.edu.","evidence":["identity"],"projectIds":[],"tags":["who","education","contact","email","uiuc"]},{"id":"berkeley-summer","kind":"FACT","text":"Donald attended University of California, Berkeley — Summer Session, EECS, Summer 2022. Summer study in electrical engineering and computer sciences.","evidence":["berkeley-summer"],"projectIds":[],"tags":["education","Berkeley","EECS","summer","2022"]},{"id":"benan-dms","kind":"FACT","text":"Shanghai Ben’an Intelligent — Research Engineer, Driver State Monitoring (DMS) (Internship), 2025 – Present · Remote. Deployed and validated real-time on-device driver-state monitoring, connecting safety-relevant alerts, low-power edge inference, and replay-driven validation. Deployed and validated eight safety-relevant detector classes: eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly. Shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics alongside model development. Designed the alert state machine with hierarchical face-loss and lens-occlusion fallbacks. Speed, ignition, and gear gating suppress false alarms in non-driving states; persisted per-camera calibration profiles support fleet-grade traceability. Fit the model pipeline to the low-power CV181x edge SoC by tightening inference cadence and per-stage compute budgets. Validated end-to-end on real cabin video and built the lab CLI for shadow-mode hard-case capture, human review, and replay-based metric reporting.","evidence":["benan-dms"],"projectIds":["driver-monitoring","dms-cv181x"],"tags":["Shanghai Ben’an Intelligent","Research Engineer, Driver State Monitoring (DMS)","On-device inference","CV181x","Python","pytest","State machines","Calibration","Replay validation","experience","Internship"]},{"id":"sjtu-research","kind":"FACT","text":"Shanghai Jiao Tong University — Autonomous Driving Research, 2021. Implemented YOLOv3 object detection, OpenCV and SLAM, and lane-following algorithms on a ROS platform for autonomous navigation. ","evidence":["sjtu-research"],"projectIds":[],"tags":["Shanghai Jiao Tong University","Autonomous Driving Research","YOLOv3","OpenCV","SLAM","ROS","Autonomous navigation","experience"]},{"id":"mckinsey-research","kind":"FACT","text":"McKinsey & Company — Part-Time Research Analyst, 2021 · Remote. Produced McKinsey-standard research reports on industrial automation and inventory management systems. ","evidence":["mckinsey-research"],"projectIds":[],"tags":["McKinsey & Company","Part-Time Research Analyst","Industrial automation","Inventory management","Research synthesis","experience"]},{"id":"current-study","kind":"FACT","text":"Donald’s current Fall 2026 coursework: ECE 385 — Digital Systems Laboratory; ECE 408 — Applied Parallel Programming; ECE 490 — Introduction to Optimization; ECE 498 — AI Systems & Engineering; IE 421 — High Frequency Trading Technology. These courses are in progress.","evidence":["academic-foundation"],"projectIds":[],"tags":["current","courses","enrollment","Fall 2026","studying","ECE 385","Digital Systems Laboratory","ECE 408","Applied Parallel Programming","ECE 490","Introduction to Optimization","ECE 498","AI Systems & Engineering","IE 421","High Frequency Trading Technology"]},{"id":"experience-overview","kind":"FACT","text":"Shanghai Ben’an Intelligent — Research Engineer, Driver State Monitoring (DMS) (Internship) (2025 – Present). Deployed and validated real-time on-device driver-state monitoring, connecting safety-relevant alerts, low-power edge inference, and replay-driven validation. Shanghai Jiao Tong University — Autonomous Driving Research (2021). Implemented YOLOv3 object detection, OpenCV and SLAM, and lane-following algorithms on a ROS platform for autonomous navigation. McKinsey & Company — Part-Time Research Analyst (2021). Produced McKinsey-standard research reports on industrial automation and inventory management systems.","evidence":["benan-dms","sjtu-research","mckinsey-research"],"projectIds":[],"tags":["experience","employment","research analyst","McKinsey","Shanghai Jiao Tong","Shanghai Ben’an Intelligent","DMS","internship"]},{"id":"academic-overview","kind":"FACT","text":"Donald’s academic foundation spans hardware & circuits; systems & algorithms; robotics & control; machine learning; mathematics & optimization; physical sciences; economics. The background page lists 24 completed UIUC core courses, five courses in progress for Fall 2026, and ten transfer or examination credit equivalents.","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","courses","classes","skills","specialties","education","transcript"]},{"id":"courses-hardware-circuits","kind":"FACT","text":"Hardware & circuits: ECE 110 Introduction to Electronics · Fall 2023; ECE 120 Introduction to Computing · Spring 2024; ECE 210 Analog Signal Processing · Spring 2024; ECE 330 Power Circuits & Electromechanics · Fall 2025; ECE 342 Electronic Circuits · Spring 2026; ECE 385 Digital Systems Laboratory · Fall 2026 (in progress).","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Hardware & circuits","Circuit measurement and modeling","Electrical circuit analysis","Sensors and motors","Electronics laboratory methods","Digital logic","Combinational and sequential circuits","Finite-state machines","Computer organization and machine language","Linear circuits and systems","Convolution and stability","Laplace and Fourier transforms","Frequency response and active filters","Power and energy","Three-phase circuits","Electromagnetic forces and torques","Electric machines and energy conversion","Transducers","Analog and digital electronic circuits","MOSFET and bipolar-transistor circuits","Amplifier analysis","Integrated-circuit design principles","ECE 110","Introduction to Electronics","ECE 120","Introduction to Computing","ECE 210","Analog Signal Processing","ECE 330","Power Circuits & Electromechanics","ECE 342","Electronic Circuits","ECE 385","Digital Systems Laboratory"]},{"id":"courses-systems-algorithms","kind":"FACT","text":"Systems & algorithms: CS 101 Introduction to Computing: Engineering & Science (exam credit); CS 124 Introduction to Computer Science I (transfer credit); CS 225 Data Structures · Spring 2025; ECE 220 Computer Systems & Programming · Fall 2024; ECE 374 Introduction to Algorithms & Models of Computation · Spring 2026; ECE 391 Computer Systems Engineering · Spring 2025; ECE 408 Applied Parallel Programming · Fall 2026 (in progress); ECE 498 AI Systems & Engineering · Fall 2026 (in progress).","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Systems & algorithms","Programming and problem solving","Elementary algorithms and data structures","Scientific and engineering computation","Fundamental programming","Computational problem solving","Introductory computing concepts","Lists, stacks, queues and trees","Object-oriented implementation","Graph and tree search","Elementary algorithm analysis","LC-3 assembly and calling conventions","C programming and pointers","Dynamic memory","Recursion and elementary data structures","Divide-and-conquer and dynamic programming","Greedy and graph algorithms","Automata and Turing machines","Reductions, undecidability and NP-completeness","Systems software","Input/output semantics","Synchronization and interrupts","Multitasking","Virtualization abstractions","CS 101","Introduction to Computing: Engineering & Science","CS 124","Introduction to Computer Science I","CS 225","Data Structures","ECE 220","Computer Systems & Programming","ECE 374","Introduction to Algorithms & Models of Computation","ECE 391","Computer Systems Engineering","ECE 408","Applied Parallel Programming","ECE 498","AI Systems & Engineering"]},{"id":"courses-robotics-control","kind":"FACT","text":"Robotics & control: ECE 470 Introduction to Robotics · Fall 2025; ECE 484 Principles of Safe Autonomy · Fall 2025; ECE 486 Control Systems · Fall 2024.","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Robotics & control","Rigid-body transformations","Forward and inverse kinematics","Motion planning and trajectories","Robotic sensing, vision and control","Autonomous-system perception","Modeling and motion planning","Control algorithms","Safety analysis and assumptions","Simulation and analysis tools","Dynamic system modeling","State-space representations","Control-system analysis and design","Computational and laboratory control methods","ECE 470","Introduction to Robotics","ECE 484","Principles of Safe Autonomy","ECE 486","Control Systems"]},{"id":"courses-machine-learning","kind":"FACT","text":"Machine learning: ECE 364 Programming Methods for Machine Learning · Spring 2025; ECE 449 Machine Learning · Spring 2026; ECE 498 Deep Generative Models · Fall 2025; CS 441 Applied Machine Learning · Spring 2026.","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Machine learning","Automatic differentiation","PyTorch implementation","Regression and classification","Neural-network programming","Clustering implementations","Supervised and generative learning","Dimensionality reduction and clustering","Neural networks","Expectation maximization","Markov decision processes and reinforcement learning","Variational inference and VAEs","Diffusion models","Generative adversarial networks","Normalizing flows","Mathematical foundations of generative modeling","Clustering","Cross-validation and bootstrap","Model selection","Neural networks and applied signal problems","ECE 364","Programming Methods for Machine Learning","ECE 449","Machine Learning","ECE 498","Deep Generative Models","CS 441","Applied Machine Learning"]},{"id":"courses-mathematics-optimization","kind":"FACT","text":"Mathematics & optimization: ECE 313 Probability with Engineering Applications · Spring 2025; MATH 213 Basic Discrete Mathematics · Fall 2023; MATH 220 Calculus (exam credit); MATH 231 Calculus II (exam credit); MATH 241 Calculus III · Fall 2023; MATH 257 Linear Algebra with Computational Applications · Spring 2024; MATH 285 Introduction to Differential Equations · Spring 2024; STAT 100 Statistics (exam credit); ECE 490 Introduction to Optimization · Fall 2026 (in progress).","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Mathematics & optimization","Probability theory","Reliability modeling","Statistical hypothesis testing","Decision-making under uncertainty","Parameter estimation","Sets, relations and functions","Counting and recurrence relations","Graphs and trees","Algorithmic reasoning","Differentiation","Integration","Fundamental theorem of calculus","Applications of single-variable calculus","Integration techniques","Polar coordinates","Conic sections","Infinite series","Partial derivatives","Multiple integrals","Vector calculus","Line and surface integrals","Linear systems and transformations","Eigenvalues and eigenvectors","Orthogonality and regression","Singular value decomposition","Computational linear algebra","Ordinary differential equations","Fourier series","Boundary-value problems","Introduction to partial differential equations","Descriptive statistics","Elementary probability","Estimation","Hypothesis testing","ECE 313","Probability with Engineering Applications","MATH 213","Basic Discrete Mathematics","MATH 220","Calculus","MATH 231","Calculus II","MATH 241","Calculus III","MATH 257","Linear Algebra with Computational Applications","MATH 285","Introduction to Differential Equations","STAT 100","Statistics","ECE 490","Introduction to Optimization"]},{"id":"courses-physical-sciences","kind":"FACT","text":"Physical sciences: CHEM 102 General Chemistry I (exam credit); CHEM 104 General Chemistry II (exam credit); PHYS 102 College Physics: Electricity, Magnetism & Modern Physics (exam credit); PHYS 211 University Physics: Mechanics · Fall 2023; PHYS 212 University Physics: Electricity & Magnetism (exam credit); PHYS 213 University Physics: Thermal Physics · Spring 2024; PHYS 214 University Physics: Quantum Physics · Spring 2024.","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Physical sciences","Atomic structure and bonding","States of matter","Stoichiometry","Chemical equilibrium","Chemical energetics","Kinetics and equilibrium","Electrochemistry","Chemistry of materials","Electric and magnetic fields","Basic circuits","Geometrical optics","Introductory modern physics","Newtonian mechanics","Work and energy","Rotational dynamics","Oscillations and waves","Electric fields and Gauss law","Capacitance and circuits","Magnetic fields and induction","Electromagnetic waves and optics","Thermodynamics","Kinetic theory","Entropy and statistical mechanics","Free energy and Boltzmann factors","Interference and diffraction","Photons and matter waves","Atomic models","Uncertainty and wave mechanics","CHEM 102","General Chemistry I","CHEM 104","General Chemistry II","PHYS 102","College Physics: Electricity, Magnetism & Modern Physics","PHYS 211","University Physics: Mechanics","PHYS 212","University Physics: Electricity & Magnetism","PHYS 213","University Physics: Thermal Physics","PHYS 214","University Physics: Quantum Physics"]},{"id":"courses-economics","kind":"FACT","text":"Economics: ECON 103 Macroeconomic Principles (exam credit); IE 421 High Frequency Trading Technology · Fall 2026 (in progress).","evidence":["academic-foundation"],"projectIds":[],"tags":["coursework","skills","Economics","Aggregate income, employment and output","Money and price levels","Monetary and fiscal policy","Inflation, unemployment and economic growth","International economics","ECON 103","Macroeconomic Principles","IE 421","High Frequency Trading Technology"]},{"id":"autonomous-driving-lab","kind":"FACT","text":"Autonomous driving lab: A ROS2/Gazebo course project connecting lane segmentation, curvature-aware speed control, and LiDAR/GPS localization. Donald implemented lane-segmentation stages, curvature-based speed control, particle-filter updates, eight-direction LiDAR support, and an offline localization parameter sweep within shared course infrastructure.","evidence":["autonomous-driving-lab-catalog-1","autonomous-driving-lab-catalog-2","autonomous-driving-lab-catalog-3","autonomous-driving-lab-catalog-4","autonomous-driving-lab-catalog-5","autonomous-driving-lab-catalog-6"],"projectIds":["autonomous-driving-lab"],"tags":["autonomous-driving-lab","Autonomous driving lab","Robotics & Autonomous Driving","Python","ROS2","Gazebo","PyTorch","SimpleENet","Particle filters","LiDAR","Vehicle control"]},{"id":"swarm-navigation","kind":"FACT","text":"Multi-robot navigation: A potential-field controller evaluated over every goal assignment for five robots in a fixed obstacle-filled circular workspace. Donald developed and tuned the potential-field controller, parameter sweeps, and trajectory visualizations for an ECE470 project using a supplied simulator.","evidence":["swarm-code","swarm-results","swarm-failure","swarm-navigation-catalog-1","swarm-navigation-catalog-2","swarm-navigation-catalog-3","swarm-navigation-catalog-4"],"projectIds":["swarm-navigation"],"tags":["swarm-navigation","Multi-robot navigation","Robotics & Autonomous Driving","Python","Control","Simulation","Potential fields","NumPy"]},{"id":"f1tenth","kind":"FACT","text":"F1Tenth lane following: A camera-to-control pipeline for a F1Tenth ROS2 platform, with confidence-aware path selection and explicit stop behavior when vision has no usable path. Donald added camera-based lane extraction, configurable image-to-robot coordinate conversion, confidence-aware path selection, and missing-path and lookahead stop behavior to an existing F1Tenth course workspace.","evidence":["control-stop","f1tenth-catalog-1","f1tenth-catalog-2","f1tenth-catalog-3","f1tenth-catalog-4"],"projectIds":["f1tenth"],"tags":["f1tenth","F1Tenth lane following","Robotics & Autonomous Driving","Python","ROS2","Computer vision","Pure pursuit","OpenCV","RealSense","Homography"]},{"id":"ur3-manipulation","kind":"FACT","text":"UR3 manipulation & vision labs: Course exercises for a UR3 arm: recursive Tower of Hanoi planning, suction and motion coordination, plus an unfinished vision-based block-manipulation extension. Student-code implementations add recursive three-block planning, state-dependent pickup and placement heights, suction feedback, colored-block detection, and image-to-world transforms to supplied course scaffolding.","evidence":["ur3-manipulation-catalog-1","ur3-manipulation-catalog-2","ur3-manipulation-catalog-3","ur3-manipulation-catalog-4"],"projectIds":["ur3-manipulation"],"tags":["ur3-manipulation","UR3 manipulation & vision labs","Robotics & Autonomous Driving","Python","ROS","UR3","OpenCV","Kinematics"]},{"id":"driver-monitoring","kind":"FACT","text":"Driver monitoring & diagnostic lab: On-device driver-state monitoring developed during Donald’s research-engineering internship at Shanghai Ben’an Intelligent: eight detector classes, vehicle-aware alerts, and replay-driven validation. Deployed and validated real-time inference for eight safety-relevant detector classes. Designed the hierarchical alert state machine and vehicle-context gating, persisted per-camera calibration profiles, and built the diagnostic lab CLI.","evidence":["benan-dms","driver-monitoring-catalog-1","driver-monitoring-catalog-2","driver-monitoring-catalog-3","driver-monitoring-catalog-4","driver-monitoring-catalog-5"],"projectIds":["driver-monitoring"],"tags":["driver-monitoring","Driver monitoring & diagnostic lab","Embedded Systems & Computer Vision","Python","MediaPipe","OpenCV","Temporal state machines","Calibration","Replay tooling"]},{"id":"dms-cv181x","kind":"FACT","text":"DMS on CV181x: A C inference pipeline for constrained hardware: sparse face detection, tracked regions, lightweight landmarks, and per-stage timing. Fit the DMS model pipeline to the low-power CV181x edge SoC during the Shanghai Ben’an Intelligent internship. Tightened inference cadence and per-stage compute budgets while adapting a supplied C service, landmark backends, and model-conversion packaging.","evidence":["benan-dms","dms-cv181x-catalog-1","dms-cv181x-catalog-2","dms-cv181x-catalog-3","dms-cv181x-catalog-4"],"projectIds":["dms-cv181x"],"tags":["dms-cv181x","DMS on CV181x","Embedded Systems & Computer Vision","C","CV181x","SCRFD","PFLD","TPU-MLIR","ONNX","Embedded inference"]},{"id":"vggt-reconstruction","kind":"FACT","text":"VGGT scene reconstruction: A local adaptation of VGGT that filters sky regions from reconstructed point clouds and adds reconstruction inspection tools. Local additions wrap upstream VGGT with ONNX sky-mask preprocessing, confidence-based point filtering, reconstruction export, and geometry inspection tools.","evidence":["vggt-reconstruction-catalog-1","vggt-reconstruction-catalog-2","vggt-reconstruction-catalog-3"],"projectIds":["vggt-reconstruction"],"tags":["vggt-reconstruction","VGGT scene reconstruction","Embedded Systems & Computer Vision","PyTorch","VGGT","ONNX Runtime","3D geometry","Viser"]},{"id":"sam3-experiments","kind":"FACT","text":"SAM 3 segmentation experiments: Prompt-based image-segmentation experiments in Meta’s SAM 3 notebook, with local environment and prompt changes. A small hands-on study modifying example prompts and Python compatibility settings in Meta’s supplied SAM 3 notebook. The model and predictor are upstream implementations.","evidence":["sam3-experiments-catalog-1","sam3-experiments-catalog-2"],"projectIds":["sam3-experiments"],"tags":["sam3-experiments","SAM 3 segmentation experiments","Embedded Systems & Computer Vision","SAM 3","PyTorch","Image segmentation","CUDA","Jupyter"]},{"id":"chronos-knn","kind":"FACT","text":"Time-series forecasting & retrieval: Time-series foundation models meet nearest-neighbor retrieval, causal evaluation, and ablations that expose prediction-to-decision failures. Donald built raw-feature retrieval baselines, model-adaptation experiments, causal PCA, and persistence-head training around upstream time-series models. The related Kronos RAG workspace is a separate local study.","evidence":["chronos-architecture","chronos-failure","chronos-regime","chronos-knn-catalog-1","chronos-knn-catalog-2","chronos-knn-catalog-3","chronos-knn-catalog-4"],"projectIds":["chronos-knn"],"tags":["chronos-knn","Time-series forecasting & retrieval","Machine Learning & Model Research","PyTorch","Qdrant","Time series","Walk-forward evaluation","Chronos","Kronos","Granite TTM","LoRA"]},{"id":"momentum-transformer-adaptation","kind":"FACT","text":"Momentum Transformer & execution learning: An extensive adaptation of the upstream Momentum Transformer research, including a PyTorch implementation and portfolio/execution reinforcement-learning layers. Donald extended Kieran Wood’s Momentum Transformer research with PyTorch modeling, execution environments, trade-delta policies, portfolio-aware agents, PPO corrections, and replay visualization.","evidence":["momentum-transformer-adaptation-catalog-1","momentum-transformer-adaptation-catalog-2","momentum-transformer-adaptation-catalog-3","momentum-transformer-adaptation-catalog-4"],"projectIds":["momentum-transformer-adaptation"],"tags":["momentum-transformer-adaptation","Momentum Transformer & execution learning","Machine Learning & Model Research","PyTorch","Temporal attention","LSTM","PPO","SAC","Execution simulation"]},{"id":"mini-torch","kind":"FACT","text":"mini_torch — autograd from NumPy: A NumPy-only neural-network training library with dynamic computation graphs, reverse-mode differentiation and a small MNIST classifier. Course implementations build dynamic computation graphs, broadcast-aware derivatives, topological backpropagation, parameterized layers, SGD, and a 784–256–128–10 classifier from NumPy.","evidence":["mini-torch-catalog-1","mini-torch-catalog-2","mini-torch-catalog-3","mini-torch-catalog-4","mini-torch-catalog-5"],"projectIds":["mini-torch"],"tags":["mini-torch","mini_torch — autograd from NumPy","Machine Learning & Model Research","Python","NumPy","Automatic differentiation","Neural networks","MNIST"]},{"id":"rl-portfolio-framework","kind":"FACT","text":"Regime-aware reinforcement learning: Recurrent PPO, portfolio environments, and related regime and fundamental-overlay studies. Each experiment retains its own evaluation boundary. Donald implemented the recurrent-policy portfolio framework and its starter. Related local TET studies add logistic overlays and fundamentals pipelines; the jump-model paper is a separate reference.","evidence":["rl-portfolio-framework-catalog-1","rl-portfolio-framework-catalog-2","rl-portfolio-framework-catalog-3","rl-portfolio-framework-catalog-4","rl-portfolio-framework-catalog-5","rl-portfolio-framework-catalog-6","rl-portfolio-framework-catalog-7","rl-portfolio-framework-catalog-8"],"projectIds":["rl-portfolio-framework"],"tags":["rl-portfolio-framework","Regime-aware reinforcement learning","Machine Learning & Model Research","PyTorch","Gymnasium","PPO","RecurrentPPO","Jump models","GMM","Streamlit","Python","Tushare","Logistic regression","Fundamental factors"]},{"id":"xtrend","kind":"FACT","text":"X-Trend reproduction: Two iterations of a few-shot forecasting reproduction, including a tensor-level correction to what attention values remember. Donald’s agent-assisted adaptation of X-Trend implements Q/K/V projections and corrects attention values to include observed context returns. Two iterations form one evolving reproduction of the upstream research.","evidence":["xtrend-values","xtrend-source","xtrend-catalog-1","xtrend-catalog-2","xtrend-catalog-3","xtrend-catalog-4","xtrend-catalog-5"],"projectIds":["xtrend"],"tags":["xtrend","X-Trend reproduction","Machine Learning & Model Research","PyTorch","Cross-attention","LSTM","Expanding windows","Gaussian processes","Sparse jump models"]},{"id":"generative-models","kind":"FACT","text":"VAE, DDPM & guided diffusion: A sequence of course studies in latent-variable generation, DDPM training and classifier-free guidance, then diffusion-based image deblurring. Course notebooks implement VAE objectives, DDPM training and sampling, classifier-free guidance, and simplified diffusion posterior sampling. Donald’s deblurring report compares six guidance scales and four measurement-noise settings.","evidence":["generative-models-catalog-1","generative-models-catalog-2","generative-models-catalog-3","generative-models-catalog-4","generative-models-catalog-5","generative-models-catalog-6"],"projectIds":["generative-models"],"tags":["generative-models","VAE, DDPM & guided diffusion","Machine Learning & Model Research","Python","PyTorch","VAE","DDPM","Classifier-free guidance","Diffusion posterior sampling","Jupyter"]},{"id":"adaptive-portfolio-learning","kind":"FACT","text":"A-share adaptive portfolio learning: Related DoubleAdapt/StockMixer, DeePM, and imitation-learning experiments in adapting models to A-share portfolio decisions. Donald adapted the upstream DeePM codebase for A-share research. Related local DoubleAdapt/StockMixer and imitation-learning implementations explore causal adaptation, feasible teachers, and reliability-based abstention.","evidence":["adaptive-portfolio-learning-catalog-1","adaptive-portfolio-learning-catalog-2","adaptive-portfolio-learning-catalog-3","adaptive-portfolio-learning-catalog-4","adaptive-portfolio-learning-catalog-5"],"projectIds":["adaptive-portfolio-learning"],"tags":["adaptive-portfolio-learning","A-share adaptive portfolio learning","Machine Learning & Model Research","PyTorch","DoubleAdapt","StockMixer","DeePM","Imitation learning","Conformal calibration","LightGBM"]},{"id":"project-genji","kind":"FACT","text":"Project Genji: temporal model evaluation: An industry-aware equity research framework whose concrete implementation centers on ingestion, feature/label contracts, purged cross-validation and supervised model training. Donald implemented feature and label contracts, time-aware cross-validation, supervised training, out-of-fold aggregation, and model persistence for industry-aware equity research.","evidence":["project-genji-catalog-1","project-genji-catalog-2","project-genji-catalog-3","project-genji-catalog-4"],"projectIds":["project-genji"],"tags":["project-genji","Project Genji: temporal model evaluation","Machine Learning & Model Research","Python","Qlib","XGBoost","Ridge regression","Purged cross-validation","Tushare"]},{"id":"symbolic-alpha-mining","kind":"FACT","text":"Symbolic alpha mining and allocation: An adapted AlphaPROBE research workspace with local point-in-time feature infrastructure and alpha-allocation additions. Local adaptations extend the upstream AlphaPROBE codebase with a point-in-time feature registry, dataset adapters, alpha-allocation modules, and targeted tests. The original algorithm and research belong to the paper authors.","evidence":["symbolic-alpha-mining-catalog-1","symbolic-alpha-mining-catalog-2","symbolic-alpha-mining-catalog-3"],"projectIds":["symbolic-alpha-mining"],"tags":["symbolic-alpha-mining","Symbolic alpha mining and allocation","Machine Learning & Model Research","Python","Symbolic expressions","GFlowNet","PPO","Point-in-time data"]},{"id":"visual-alpha-research","kind":"FACT","text":"A-share representation and attention-model adaptations: Chart-image CNN/ViT, market-guided attention, and cross-sectional representation studies in one model-research family. Local implementations and adaptations explore chart-image CNNs/ViTs, MASTER adapters, and Attention Factors. Each original paper and model remains credited to its authors.","evidence":["visual-alpha-research-catalog-1","visual-alpha-research-catalog-2","visual-alpha-research-catalog-3","visual-alpha-research-catalog-4","visual-alpha-research-catalog-5","visual-alpha-research-catalog-6","visual-alpha-research-catalog-7"],"projectIds":["visual-alpha-research"],"tags":["visual-alpha-research","A-share representation and attention-model adaptations","Machine Learning & Model Research","PyTorch","CNN","Vision Transformer","MASTER","t-SNE","A-share data","Attention factors","Statistical arbitrage","Point-in-time panels"]},{"id":"rotation-learning","kind":"FACT","text":"Leader-follower rotation and event ranking: Graph, GNN, and HMM experiments investigate which stocks recover after a market leader’s limit-up streak ends. Local research implementations compare a signed temporal graph, a dynamic latent GNN, and a Student-t HMM choice model, with separate data and evaluation contracts.","evidence":["rotation-learning-catalog-1","rotation-learning-catalog-2","rotation-learning-catalog-3","rotation-learning-catalog-4","rotation-learning-catalog-5"],"projectIds":["rotation-learning"],"tags":["rotation-learning","Leader-follower rotation and event ranking","Machine Learning & Model Research","GNN","Student-t HMM","Temporal encoders","Plackett-Luce","Calibration","Event ranking"]},{"id":"cnn-ablation-harness","kind":"FACT","text":"CIFAR-10 experiment harness: A modular reproduction of the PyTorch CIFAR-10 tutorial that turns seed, augmentation, epoch and channel-width changes into selectable experiment configurations. Local modules separate configuration, seeding, data transforms, model, training and evaluation, with CLI experiment flags and lightweight wiring/smoke tests.","evidence":["cnn-ablation-harness-catalog-1","cnn-ablation-harness-catalog-2","cnn-ablation-harness-catalog-3","cnn-ablation-harness-catalog-4","cnn-ablation-harness-catalog-5"],"projectIds":["cnn-ablation-harness"],"tags":["cnn-ablation-harness","CIFAR-10 experiment harness","Machine Learning & Model Research","Python","PyTorch","CIFAR-10","Experiment configuration"]},{"id":"upstream-ai-reference-library","kind":"FACT","text":"Forecasting & market-agent reference library: CAMEF and the Kalshi AI bot: upstream codebases retained as reference foundations, with original authorship clearly distinguished. Reference study of upstream CAMEF and Ryan Frigo’s Kalshi AI bot. These clean checkouts are original authors’ work, with no local implementation contribution claimed.","evidence":["upstream-ai-reference-library-catalog-1","upstream-ai-reference-library-catalog-2","upstream-ai-reference-library-catalog-3","upstream-ai-reference-library-catalog-4","upstream-ai-reference-library-catalog-5"],"projectIds":["upstream-ai-reference-library"],"tags":["upstream-ai-reference-library","Forecasting & market-agent reference library","Machine Learning & Model Research","Multimodal learning","Causal forecasting","Counterfactual augmentation","Python","LLM applications","Prediction markets"]},{"id":"local-qwen-inference","kind":"FACT","text":"Local Qwen inference: Self-hosted Qwen workflows with controlled request budgets, reproducible decoding, structured decisions, and explicit inference-failure handling. Application-side controls wrap upstream Qwen/vLLM with durable request budgets, fixed decoding settings, runtime metadata, and structured outputs. Donald also contributed private-serving configuration in the event-ending project.","evidence":["local-qwen-inference-catalog-1","local-qwen-inference-catalog-2","local-qwen-inference-catalog-3","local-qwen-inference-catalog-4"],"projectIds":["local-qwen-inference"],"tags":["local-qwen-inference","Local Qwen inference","Local LLM Inference & Agents","Qwen","vLLM","OpenAI-compatible HTTP","JSON Schema","Python","fcntl","fsync","Quantization","PyTorch"]},{"id":"zhengmind-yagni","kind":"FACT","text":"ZhengMindYAGNI — agent orchestration: An AI-work coordinator built around Git, markdown work nodes, a small state machine, and a CLI that owns mutations. Donald simplified the state machine, removed unused machinery, added reusable lessons, and made agent-loop progress observable. The implementation keeps state changes behind a CLI and uses exclusive claims for ready work.","evidence":["zhengmind-yagni-catalog-1","zhengmind-yagni-catalog-2","zhengmind-yagni-catalog-3","zhengmind-yagni-catalog-4","zhengmind-yagni-catalog-5","zhengmind-yagni-catalog-6"],"projectIds":["zhengmind-yagni"],"tags":["zhengmind-yagni","ZhengMindYAGNI — agent orchestration","Local LLM Inference & Agents","Go","Git","Markdown","JSONL","File locks","CLI","HTTP","Claude CLI"]},{"id":"vta-qwen-adaptation","kind":"FACT","text":"VTA — local Qwen adaptation: A local adaptation of an upstream financial time-series LLM pipeline to Qwen3.5, spanning inference compatibility and staged GRPO/LoRA/SFT training code. Local changes adapt the upstream chen-jan pipeline to Qwen3.5 chat and processor conventions, distinguish model-loading paths, and add an environment-specific dependency profile. The original algorithms and training workflow remain upstream work.","evidence":["vta-qwen-adaptation-catalog-1","vta-qwen-adaptation-catalog-2","vta-qwen-adaptation-catalog-3","vta-qwen-adaptation-catalog-4","vta-qwen-adaptation-catalog-5"],"projectIds":["vta-qwen-adaptation"],"tags":["vta-qwen-adaptation","VTA — local Qwen adaptation","Local LLM Inference & Agents","Qwen3.5","Unsloth","PyTorch","Transformers","TRL","LoRA","GRPO","SFT","CUDA","Python"]},{"id":"qwen-market-ending","kind":"FACT","text":"Event-ending inference with Qwen: A local-Qwen decision pipeline asks whether the real-world event behind a prediction market is likely to end within a chosen horizon. A shared implementation combines exchange facts, companion-market context, local-model calls, and comparison with a rule baseline. Donald contributed configurable private serving; the source also records jd contributions.","evidence":["qwen-market-ending-catalog-1","qwen-market-ending-catalog-2","qwen-market-ending-catalog-3","qwen-market-ending-catalog-4"],"projectIds":["qwen-market-ending"],"tags":["qwen-market-ending","Event-ending inference with Qwen","Local LLM Inference & Agents","Python standard library","Qwen","vLLM","HTTP APIs","JSON","Concurrent futures","Evaluation"]},{"id":"semantic-research-agents","kind":"FACT","text":"Semantic research agents: A-share research workflows that turn textual evidence into structured long-only decisions, with accounting and governance checks. Donald-attributed commits extend the upstream TradingAgents and ATLAS frameworks with specialized A-share research workflows. A related standalone semantic engine implements confirmation signals, vetoes and constrained portfolios.","evidence":["semantic-research-agents-catalog-1","semantic-research-agents-catalog-2","semantic-research-agents-catalog-3","semantic-research-agents-catalog-4"],"projectIds":["semantic-research-agents"],"tags":["semantic-research-agents","Semantic research agents","Local LLM Inference & Agents","Python","LLM agents","Structured outputs","TradingAgents","ATLAS","Tushare"]},{"id":"agent-context-blueprint","kind":"FACT","text":"Typed handoffs for coding agents: A reusable agent workflow that separates planning, interface skeletons, implementation, and integration through typed patch artifacts and deterministic hooks. Donald built hook-driven integration, branch policy, durable review outputs, and documentation around typed handoffs between planning, implementation, and review roles.","evidence":["agent-context-blueprint-catalog-1","agent-context-blueprint-catalog-2","agent-context-blueprint-catalog-3","agent-context-blueprint-catalog-4"],"projectIds":["agent-context-blueprint"],"tags":["agent-context-blueprint","Typed handoffs for coding agents","Local LLM Inference & Agents","Python","Shell","JSON Schema","Git","Claude Code hooks","Codex","Agent orchestration"]},{"id":"claude-code-autopilot","kind":"FACT","text":"Claude Code Autopilot: An installation and activation layer combining upstream coding skills, specialized agents, and hooks into a reusable project toolkit. Donald built the installer and refined session-start and review behavior around upstream obra/superpowers and diet103 skills. Those integrated skills and agents remain credited to their authors.","evidence":["claude-code-autopilot-catalog-1","claude-code-autopilot-catalog-2","claude-code-autopilot-catalog-3","claude-code-autopilot-catalog-4"],"projectIds":["claude-code-autopilot"],"tags":["claude-code-autopilot","Claude Code Autopilot","Local LLM Inference & Agents","Shell","Python","TypeScript","Claude Code","Hooks","Codex","Skills"]},{"id":"zhengmind-foundation","kind":"FACT","text":"ZhengMind architecture study: A provider-neutral task-control architecture studied alongside the later ZhengMindYAGNI implementation. An architectural study of the ZhengMind prototype credited to yq77zs73/Zheng Shen. It is separate from Donald’s independently attributed ZhengMindYAGNI implementation.","evidence":["zhengmind-foundation-catalog-1","zhengmind-foundation-catalog-2","zhengmind-foundation-catalog-3","zhengmind-foundation-catalog-4","zhengmind-foundation-catalog-5"],"projectIds":["zhengmind-foundation"],"tags":["zhengmind-foundation","ZhengMind architecture study","Local LLM Inference & Agents","Python","FastAPI","SQLite","Jinja2","Task graphs","Scoped tokens","Audit trails"]},{"id":"serena-foundation","kind":"FACT","text":"Serena semantic tooling reference: An upstream reference for symbol-aware code retrieval and editing through language servers and MCP. Reference study of the upstream Serena toolkit; no original implementation or local integration is claimed.","evidence":["serena-foundation-catalog-1","serena-foundation-catalog-2"],"projectIds":["serena-foundation"],"tags":["serena-foundation","Serena semantic tooling reference","Local LLM Inference & Agents","Python","MCP","Language Server Protocol","Semantic code tools"]},{"id":"isafe-sop","kind":"FACT","text":"ISAFE / WITNESS — SOP verification: An assembly-inspection prototype that tracks procedure steps, flags skips and ordering errors, and attaches video evidence for review. Donald built the Python backend, integrated perception and adjudication components, repaired conformance logic, and iterated the review interface. Perception and vision-language models are upstream integrations.","evidence":["isafe-sop-catalog-1","isafe-sop-catalog-2","isafe-sop-catalog-3","isafe-sop-catalog-4","isafe-sop-catalog-5","isafe-sop-catalog-6"],"projectIds":["isafe-sop"],"tags":["isafe-sop","ISAFE / WITNESS — SOP verification","Software Systems & Automation","Python","FastAPI","Pydantic","NumPy","Soft-DTW","DAG/FSM","Qwen2.5-VL","PyTorch","bitsandbytes","DINOv2","FAISS","SQLite","PostgreSQL","MQTT","WebSocket"]},{"id":"cpp-goroutine","kind":"FACT","text":"A Go-style runtime in C++: Cooperative scheduling, epoll-backed I/O, timers, and synchronization in C++17. Waiting tasks yield instead of blocking a worker. Implements a C++17 scheduler with guard-page stacks, epoll-backed I/O, timers, libc hooks, joinable tasks, and synchronization that yields cooperatively.","evidence":["cpp-goroutine-catalog-1","cpp-goroutine-catalog-2","cpp-goroutine-catalog-3","cpp-goroutine-catalog-4","cpp-goroutine-catalog-5","cpp-goroutine-catalog-6","cpp-goroutine-catalog-7"],"projectIds":["cpp-goroutine"],"tags":["cpp-goroutine","A Go-style runtime in C++","Software Systems & Automation","C++17","Linux","ucontext","epoll","eventfd","pthread","CMake"]},{"id":"applypilot-adaptation","kind":"FACT","text":"ApplyPilot workflow adaptation: A local ApplyPilot adaptation adding a Claude CLI provider and crash-resilient parallel scoring. Extends Pickle-Pixel/Henry Muhiar’s ApplyPilot with an isolated Claude CLI adapter, retries, parallel scoring, and incremental SQLite commits. The underlying application agent is upstream work.","evidence":["applypilot-adaptation-catalog-1","applypilot-adaptation-catalog-2","applypilot-adaptation-catalog-3","applypilot-adaptation-catalog-4"],"projectIds":["applypilot-adaptation"],"tags":["applypilot-adaptation","ApplyPilot workflow adaptation","Software Systems & Automation","Python","httpx","Claude CLI","ThreadPoolExecutor","SQLite","Subprocess"]},{"id":"invoice-authentication","kind":"FACT","text":"Full-stack authentication study: Alternative native-authentication implementations inside a supplied React/GraphQL invoice application, with JWT access tokens, rotating refresh sessions and cookie-based browser integration. Local variants extend David Andrea/djblackett’s invoice app with bcrypt password checks, signed access tokens, hashed refresh records, rotation/reuse handling, and authentication routes and services.","evidence":["invoice-authentication-catalog-1","invoice-authentication-catalog-2","invoice-authentication-catalog-3","invoice-authentication-catalog-4","invoice-authentication-catalog-5"],"projectIds":["invoice-authentication"],"tags":["invoice-authentication","Full-stack authentication study","Software Systems & Automation","TypeScript","React","Express","GraphQL","Prisma","PostgreSQL","JWT","Vitest"]},{"id":"cuda-introduction","kind":"FACT","text":"CUDA development environment study: Introductory CUDA course material for compiling a supplied device-query program and running it through Slurm. An introductory course reference using the supplied CUDA device-query program, compiler setup, and Slurm job files. No custom GPU kernel is claimed.","evidence":["cuda-introduction-catalog-1","cuda-introduction-catalog-2","cuda-introduction-catalog-3"],"projectIds":["cuda-introduction"],"tags":["cuda-introduction","CUDA development environment study","Software Systems & Automation","CUDA","C++","Slurm"]},{"id":"blockchain-principles","kind":"FACT","text":"Blockchain principles reading collection: A local collection of blockchain course slides and notes. A reading reference composed of supplied blockchain course slides and notes, rather than an original implementation.","evidence":["blockchain-principles-catalog-1","blockchain-principles-catalog-2"],"projectIds":["blockchain-principles"],"tags":["blockchain-principles","Blockchain principles reading collection","Software Systems & Automation","Distributed systems","Blockchain coursework"]},{"id":"kalshibot","kind":"FACT","text":"Prediction-market execution systems: Order reconciliation, execution controls, queue-aware research, and an independently packaged refactor of a prediction-market system. Donald’s agent-assisted independent refactor separates execution, tape, backtesting, research, and dashboards, with explicit runtime paths and write controls. Earlier system and research snapshots retain shared-source attribution.","evidence":["kalshi-refactor","kalshi-ack","kalshi-checks","kalshibot-catalog-1","kalshibot-catalog-2","kalshibot-catalog-3","kalshibot-catalog-4","kalshibot-catalog-5"],"projectIds":["kalshibot"],"tags":["kalshibot","Prediction-market execution systems","Quantitative Finance & Markets","Python","Reconciliation","Event-driven systems","Offline testing","Queue-aware replay","Qwen","OpenAI-compatible inference"]},{"id":"statistical-arbitrage","kind":"FACT","text":"Statistical arbitrage & module discovery: Graph-conditioned residuals, pair selection, and a constrained search harness for testing alternative research modules. Donald implemented graph, residual, expected-net-alpha, pair-selection, and execution layers, plus a bounded task/evaluator pack for exploring alternative research modules.","evidence":["statistical-arbitrage-catalog-1","statistical-arbitrage-catalog-2","statistical-arbitrage-catalog-3","statistical-arbitrage-catalog-4"],"projectIds":["statistical-arbitrage"],"tags":["statistical-arbitrage","Statistical arbitrage & module discovery","Quantitative Finance & Markets","Python","Graphs","VECM","Statistical arbitrage","SkyDiscover","Deterministic evaluators"]},{"id":"options-variance","kind":"FACT","text":"Options & variance-risk-premium research: Options research and execution tooling spanning implied-versus-realized volatility, short-strangle protocols, and an earnings-event variance decomposition study. Donald implemented pricing, option-chain selection, strategy/backtest modules, and an IBKR API layer. A related earnings-event study separates jump variance from diffusion variance.","evidence":["options-variance-catalog-1","options-variance-catalog-2","options-variance-catalog-3","options-variance-catalog-4"],"projectIds":["options-variance"],"tags":["options-variance","Options & variance-risk-premium research","Quantitative Finance & Markets","Python","IBKR API","Options pricing","Variance decomposition","Walk-forward research"]},{"id":"lppls-research-cockpits","kind":"FACT","text":"LPPLS market-state research and interactive cockpits: LPPLS bubble detection, an inspectable research cockpit, and related market-mechanism studies with frozen-history checks and rollback controls. Local implementations connect LPPLS fitting, meta-labeling, validation, backtests, and two research cockpits. The A-share service supports dated research and historical reconciliation; the global monitor fits trailing windows on indices and ETFs.","evidence":["lppls-research-cockpits-catalog-1","lppls-research-cockpits-catalog-2","lppls-research-cockpits-catalog-3","lppls-research-cockpits-catalog-4","lppls-research-cockpits-catalog-5","lppls-research-cockpits-catalog-6","lppls-research-cockpits-catalog-7","lppls-research-cockpits-catalog-8","lppls-research-cockpits-catalog-9"],"projectIds":["lppls-research-cockpits"],"tags":["lppls-research-cockpits","LPPLS market-state research and interactive cockpits","Quantitative Finance & Markets","Python","LPPLS","FastAPI","Purged cross-validation","Isotonic calibration","Parquet","Interactive dashboards"]},{"id":"zero-human-hedge","kind":"FACT","text":"Zero Human Hedge — paper research operations: An integrated paper-research stack connecting a minute-data lab, portfolio optimizer, Nautilus parity backtest, observability and an agent control room. Donald connected data aggregation, feature research, portfolio optimization, parity backtesting, and report synchronization through CLI and application integrations.","evidence":["zero-human-hedge-catalog-1","zero-human-hedge-catalog-2","zero-human-hedge-catalog-3"],"projectIds":["zero-human-hedge"],"tags":["zero-human-hedge","Zero Human Hedge — paper research operations","Quantitative Finance & Markets","Python","DuckDB","NautilusTrader","Paperclip","Docker","Prometheus","Grafana"]},{"id":"mention-market-model","kind":"FACT","text":"World Cup announcer-mention modeling: A distinct event-market study combining broadcaster transcripts, settlement labels, crew effects and mention-time approximations. A standalone local study implements transcript-corpus ingestion, settlement-label processing, mention-time approximations, and hierarchical broadcaster/team effects, separate from the upstream Kalshi bot.","evidence":["mention-market-model-catalog-1","mention-market-model-catalog-2","mention-market-model-catalog-3"],"projectIds":["mention-market-model"],"tags":["mention-market-model","World Cup announcer-mention modeling","Quantitative Finance & Markets","Python","NLP corpora","Empirical Bayes","Beta-binomial models","Calibration"]},{"id":"factor-replication","kind":"FACT","text":"A-share factor replication and evaluation: Asset-pricing factor replication, index-enhancement experiments, and related upstream data infrastructure. Local factor implementations and evaluation harnesses adapt published predictor definitions to A-share availability and trading constraints. Hikyuu remains upstream infrastructure with a scoped local import change.","evidence":["factor-replication-catalog-1","factor-replication-catalog-2","factor-replication-catalog-3","factor-replication-catalog-4","factor-replication-catalog-5"],"projectIds":["factor-replication"],"tags":["factor-replication","A-share factor replication and evaluation","Quantitative Finance & Markets","Python","Tushare","Asset pricing","Factor models","PIT universes","Statistical inference","C++","HDF5","Market-data ingestion"]},{"id":"mean-reversion-system","kind":"FACT","text":"A-share reversal with executable constraints: A short-horizon reversal system with clustering, market-specific tradeability, portfolio construction and futures hedging. Donald implemented alpha, universe, feature, portfolio, execution, and backtest modules, including clustering-enhanced reversal signals and market-specific tradeability checks.","evidence":["mean-reversion-system-catalog-1","mean-reversion-system-catalog-2","mean-reversion-system-catalog-3"],"projectIds":["mean-reversion-system"],"tags":["mean-reversion-system","A-share reversal with executable constraints","Quantitative Finance & Markets","Python","Clustering","Mean reversion","Portfolio optimization","Execution simulation"]},{"id":"market-signal-ingestion","kind":"FACT","text":"Market-signal ingestion and replay: A multi-source signal-ingestion, replay and alerting layer for prediction-market research. Donald built source adapters, stream capture and replay, downstream alert sinks, and a live-source dashboard.","evidence":["market-signal-ingestion-catalog-1","market-signal-ingestion-catalog-2","market-signal-ingestion-catalog-3"],"projectIds":["market-signal-ingestion"],"tags":["market-signal-ingestion","Market-signal ingestion and replay","Quantitative Finance & Markets","Python","httpx","WebSockets","REST APIs","Replay","Dashboards"]},{"id":"technical-indicator-evaluation","kind":"FACT","text":"Technical-indicator evaluation workbench: A repeatable multi-fold evaluation and dashboard system for technical signals on US minute bars. Local implementations define a common indicator contract and connect discovery, isolated evaluation, failure replay, checkpointing, and serving.","evidence":["technical-indicator-evaluation-catalog-1","technical-indicator-evaluation-catalog-2"],"projectIds":["technical-indicator-evaluation"],"tags":["technical-indicator-evaluation","Technical-indicator evaluation workbench","Quantitative Finance & Markets","Python","pandas","NumPy","Streamlit","Parquet","Factor evaluation"]},{"id":"market-mechanism-studies","kind":"FACT","text":"A-share event, flow and intervention studies: Studies of market bottoms, public flow disclosures, IPOs, and observable policy-support signals, with rejected hypotheses retained. Local experiment harnesses, reports, and a Streamlit scoring engine turn event, disclosure, and policy-support hypotheses into measurable tests.","evidence":["market-mechanism-studies-catalog-1","market-mechanism-studies-catalog-2","market-mechanism-studies-catalog-3","market-mechanism-studies-catalog-4","market-mechanism-studies-catalog-5","market-mechanism-studies-catalog-6"],"projectIds":["market-mechanism-studies"],"tags":["market-mechanism-studies","A-share event, flow and intervention studies","Quantitative Finance & Markets","Python","Event studies","Tushare","Time-series analysis","Hypothesis testing","Streamlit","Robust statistics","Interpretable scoring"]},{"id":"finance-reading-and-falsification","kind":"FACT","text":"Trading literature and falsification notebooks: Research notes that turn trading literature and informal claims into testable rules, experiment plans, and critical assessments. Research artifacts translate trading literature into a long-only blueprint, testable mechanical claims, experiment plans, and critical assessments. They are not presented as Donald’s verified first-person writing.","evidence":["finance-reading-and-falsification-catalog-1","finance-reading-and-falsification-catalog-2","finance-reading-and-falsification-catalog-3","finance-reading-and-falsification-catalog-4"],"projectIds":["finance-reading-and-falsification"],"tags":["finance-reading-and-falsification","Trading literature and falsification notebooks","Quantitative Finance & Markets","Literature synthesis","Experiment design","Falsification","Quantitative research"]},{"id":"csc-tactics","kind":"FACT","text":"CSC:tactics: A Chinese soccer tactics teaching application with 52 situational challenges, 11 roles, three viewpoints and a 90-minute training plan containing 43 segments. The project combines Chinese teaching content, 52 situational challenges, 11 selectable roles, three viewpoints, and a structured training plan using shared Three.js/Canvas scene logic.","evidence":["soccer-content","soccer-lifecycle","soccer-reflection","csc-tactics-catalog-1","csc-tactics-catalog-2","csc-tactics-catalog-3","csc-tactics-catalog-4"],"projectIds":["csc-tactics"],"tags":["csc-tactics","CSC:tactics","Interactive Tools, Learning & Writing","JavaScript","Three.js","Canvas","Vite","Interactive learning"]},{"id":"youtube-knowledge","kind":"FACT","text":"YouTube Knowledge: A Chinese-first reading and exploration application that turns video transcripts into staged summaries, source-linked ideas, and creator-level knowledge graphs. Donald built transcript ingestion, streaming APIs, source-linked distillation, creator knowledge graphs, and query/feed views, then improved concurrency and mobile behavior.","evidence":["youtube-knowledge-catalog-1","youtube-knowledge-catalog-2","youtube-knowledge-catalog-3","youtube-knowledge-catalog-4","youtube-knowledge-catalog-5","youtube-knowledge-catalog-6"],"projectIds":["youtube-knowledge"],"tags":["youtube-knowledge","YouTube Knowledge","Interactive Tools, Learning & Writing","Python","Claude CLI","NDJSON","Threads","Atomic file replacement","JSON","YouTube transcripts","JavaScript"]},{"id":"piano-midi","kind":"FACT","text":"Piano video to MIDI: A web pipeline converting YouTube/Bilibili piano audio into editable MIDI, with optional timing/dynamics cleanup and a saved library. Donald built video ingestion, job orchestration, library deduplication, and MIDI postprocessing around upstream Transkun transcription.","evidence":["piano-pipeline","piano-midi-catalog-1","piano-midi-catalog-2","piano-midi-catalog-3","piano-midi-catalog-4"],"projectIds":["piano-midi"],"tags":["piano-midi","Piano video to MIDI","Interactive Tools, Learning & Writing","Python","Flask","Transkun","MIDI","PyTorch","pretty-midi","yt-dlp"]},{"id":"browser-repl","kind":"FACT","text":"Sandbox: browser code experiments: Two related Vue/CodeMirror implementations of a browser-only JavaScript/Python editor with execution, console output, saved files and export. Local Vue/CodeMirror variants implement editing, language switching, file persistence, console output, and export. One uses an iframe/main-thread runtime; the other runs JavaScript and Python in terminable workers.","evidence":["browser-repl-catalog-1","browser-repl-catalog-2","browser-repl-catalog-3","browser-repl-catalog-4"],"projectIds":["browser-repl"],"tags":["browser-repl","Sandbox: browser code experiments","Interactive Tools, Learning & Writing","Vue","CodeMirror","Vite","Web Workers","Pyodide","WebAssembly"]},{"id":"daily-news-terminal","kind":"FACT","text":"Daily News Vision Terminal: A dockerized live-information interface combining news signals, prediction-market context, public trade activity, and ADS-B flight overlays. The local implementation normalizes public feeds, scores text-to-market relevance, manages keyword alerts, and streams updates to a globe and dense news interface.","evidence":["daily-news-terminal-catalog-1","daily-news-terminal-catalog-2","daily-news-terminal-catalog-3","daily-news-terminal-catalog-4"],"projectIds":["daily-news-terminal"],"tags":["daily-news-terminal","Daily News Vision Terminal","Interactive Tools, Learning & Writing","Node.js","JavaScript","WebSocket","Docker","GDELT","Polymarket public APIs","OpenSky ADS-B"]},{"id":"video-transcriber-adaptation","kind":"FACT","text":"Video transcription workflow adaptation: Local adaptations to an upstream subtitle-first transcription and summarization application, including a different model client, GPU transcription and platform-specific retrieval handling. Local changes extend wendy7756’s AI-Video-Transcriber with a Claude Agent SDK adapter, shared model configuration, CUDA float16 Faster-Whisper, and Bilibili retrieval handling.","evidence":["video-transcriber-adaptation-catalog-1","video-transcriber-adaptation-catalog-2","video-transcriber-adaptation-catalog-3","video-transcriber-adaptation-catalog-4","video-transcriber-adaptation-catalog-5"],"projectIds":["video-transcriber-adaptation"],"tags":["video-transcriber-adaptation","Video transcription workflow adaptation","Interactive Tools, Learning & Writing","Python","FastAPI","Faster-Whisper","CUDA","yt-dlp","Claude Agent SDK"]},{"id":"personal-website","kind":"FACT","text":"A searchable engineering notebook: This site: a field-based project catalog and a local-Qwen guide that answers from reviewed claims and linked evidence. Donald directed the information model and design, including the discipline-based reorganization. The implementation was built with an AI coding agent: shared structured content, ranked project families, a local-Qwen evidence guide and immutable shared answers.","evidence":["personal-website-catalog-1","personal-website-catalog-2","personal-website-catalog-3"],"projectIds":["personal-website"],"tags":["personal-website","A searchable engineering notebook","Interactive Tools, Learning & Writing","React","TypeScript","Cloudflare","Local Qwen","Semantic HTML"]},{"id":"technology-society-writing","kind":"FACT","text":"Cars, identity & technological change: A candidate-attributed course essay examines American Graffiti through histories of mobility, vehicle identity and the social consequences of transport technology. Donald’s course essay connects American Graffiti to historical readings on American mobility, the Yugo, decorated trucks, and railroads.","evidence":["technology-society-writing-catalog-1","technology-society-writing-catalog-2"],"projectIds":["technology-society-writing"],"tags":["technology-society-writing","Cars, identity & technological change","Interactive Tools, Learning & Writing","Academic writing","Technology history","LaTeX"]},{"id":"illinix-391","kind":"FACT","text":"Illinix 391: A Unix-like kernel in C and RISC-V assembly, from virtual memory to an interactive shell. Architected a kernel from scratch with virtual memory, fork/exec, 15+ system calls, VIRTIO block drivers, and a custom KTFS filesystem.","evidence":["illinix-391-catalog-1"],"projectIds":["illinix-391"],"tags":["illinix-391","Illinix 391","Software Systems & Automation","C","RISC-V Assembly","Virtual memory","Operating systems","VIRTIO","KTFS"]},{"id":"reaction-wheel-pendulum","kind":"FACT","text":"Reaction Wheel Pendulum: Real-time stabilization of an inherently unstable pendulum using a reaction wheel. Modeled the system using Lagrangian dynamics and implemented three-state feedback with a Luenberger observer and friction compensation.","evidence":["reaction-wheel-pendulum-catalog-1"],"projectIds":["reaction-wheel-pendulum"],"tags":["reaction-wheel-pendulum","Reaction Wheel Pendulum","Robotics & Autonomous Driving","Control systems","Lagrangian dynamics","State feedback","Luenberger observer","Wincon"]},{"id":"twentyfourpoints","kind":"FACT","text":"twentyfourpoints.com: A real-time multiplayer math arena, built and launched in 72 hours. Built the React/TypeScript client, Node.js and Socket.io multiplayer backend, Supabase authentication and ELO rankings, and a custom mathematical game engine.","evidence":["twentyfourpoints-catalog-1"],"projectIds":["twentyfourpoints"],"tags":["twentyfourpoints","twentyfourpoints.com","Interactive Tools, Learning & Writing","React","TypeScript","Node.js","Socket.io","Supabase","ELO"]},{"id":"autonomous-driving-lab-details","kind":"FACT","text":"Autonomous driving lab: Lane masks transformed into bird-eye geometry and fitted lane curves; waypoint lookahead and turn anticipation for speed selection; particle likelihoods with finite-weight safeguards, systematic resampling and GPS reseeding under low effective sample size; eight-direction LiDAR support; offline generated motion/noise trials measuring position RMSE, heading RMSE, convergence and ESS. An integrated set of perception, control, and localization modules, with training checkpoints and simulation experiments that compare position error, heading error, convergence, and effective sample size.","evidence":["autonomous-driving-lab-catalog-1","autonomous-driving-lab-catalog-2","autonomous-driving-lab-catalog-3","autonomous-driving-lab-catalog-4","autonomous-driving-lab-catalog-5","autonomous-driving-lab-catalog-6"],"projectIds":["autonomous-driving-lab"],"tags":["autonomous-driving-lab","Autonomous driving lab","implementation","technical depth","architecture","result","Python","ROS2","Gazebo","PyTorch","SimpleENet","Particle filters","LiDAR","Vehicle control"]},{"id":"autonomous-driving-lab-limits","kind":"FACT","text":"Autonomous driving lab — evidence boundaries: Shared course/team workspace; do not attribute the simulator, model backbone or all source to Donald. Configured rates and speeds are not measured physical performance. Offline tuning uses synthetic sequences and cannot substitute for physical vehicle evaluation. MP0 contains a preliminary decision-rule study with remaining scaffolding; it is not a separately verified autonomous system.","evidence":["autonomous-driving-lab-catalog-1","autonomous-driving-lab-catalog-2","autonomous-driving-lab-catalog-3","autonomous-driving-lab-catalog-4","autonomous-driving-lab-catalog-5","autonomous-driving-lab-catalog-6"],"projectIds":["autonomous-driving-lab"],"tags":["autonomous-driving-lab","Autonomous driving lab","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"swarm-navigation-details","kind":"FACT","text":"Multi-robot navigation: Too much goal attraction can overpower repulsion. Local minima and orbiting can prevent completion. All 120 goal assignments completed in the saved five-robot simulation suite, averaging 8,434 simulation steps. The result applies to that fixed simulator and obstacle configuration.","evidence":["swarm-code","swarm-results","swarm-failure","swarm-navigation-catalog-1","swarm-navigation-catalog-2","swarm-navigation-catalog-3","swarm-navigation-catalog-4"],"projectIds":["swarm-navigation"],"tags":["swarm-navigation","Multi-robot navigation","implementation","technical depth","architecture","result","Python","Control","Simulation","Potential fields","NumPy"]},{"id":"swarm-navigation-limits","kind":"FACT","text":"Multi-robot navigation — evidence boundaries: Fixed course simulation suite, not a general convergence guarantee or physical-robot result. Prose report min/max differ from the saved CSV; use only checked 120/120 and mean. Excessive attraction caused collisions; nonlinear attraction addressed orbiting/local minima. The controller remains reactive and does not anticipate future conflicts.","evidence":["swarm-code","swarm-results","swarm-failure","swarm-navigation-catalog-1","swarm-navigation-catalog-2","swarm-navigation-catalog-3","swarm-navigation-catalog-4"],"projectIds":["swarm-navigation"],"tags":["swarm-navigation","Multi-robot navigation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"f1tenth-details","kind":"FACT","text":"F1Tenth lane following: Pixel geometry must become robot-frame geometry. Confidence and age matter when choosing a path. A perception-to-control implementation that selects paths by confidence and freshness, then stops camera-only control when no usable path remains.","evidence":["control-stop","f1tenth-catalog-1","f1tenth-catalog-2","f1tenth-catalog-3","f1tenth-catalog-4"],"projectIds":["f1tenth"],"tags":["f1tenth","F1Tenth lane following","implementation","technical depth","architecture","result","Python","ROS2","Computer vision","Pure pursuit","OpenCV","RealSense","Homography"]},{"id":"f1tenth-limits","kind":"FACT","text":"F1Tenth lane following — evidence boundaries: Existing UIUC Robotics/F1Tenth foundation with multiple contributors. Nominal camera rates, control rates and speeds are configurations rather than measurements. No claim of a completed physical track trial or real-world reliability. Physical track performance is still unverified.","evidence":["control-stop","f1tenth-catalog-1","f1tenth-catalog-2","f1tenth-catalog-3","f1tenth-catalog-4"],"projectIds":["f1tenth"],"tags":["f1tenth","F1Tenth lane following","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"ur3-manipulation-details","kind":"FACT","text":"UR3 manipulation & vision labs: ROS command/feedback loops; recursive task planning; arm and vacuum control; calibrated coordinate transforms; morphological cleanup and blob filtering for colored blocks. The Hanoi implementation combines a recursive planner with calibrated arm positions. The vision extension includes blob extraction but still needs its kinematics and calibration/goal placeholders completed.","evidence":["ur3-manipulation-catalog-1","ur3-manipulation-catalog-2","ur3-manipulation-catalog-3","ur3-manipulation-catalog-4"],"projectIds":["ur3-manipulation"],"tags":["ur3-manipulation","UR3 manipulation & vision labs","implementation","technical depth","architecture","result","Python","ROS","UR3","OpenCV","Kinematics"]},{"id":"ur3-manipulation-limits","kind":"FACT","text":"UR3 manipulation & vision labs — evidence boundaries: No physical execution success or author-by-line provenance verified. Lab5 Get_MS returns identity/zero scaffolding and inverse kinematics is unimplemented; do not present it as completed vision pick-and-place. Course drivers and starter code are upstream.","evidence":["ur3-manipulation-catalog-1","ur3-manipulation-catalog-2","ur3-manipulation-catalog-3","ur3-manipulation-catalog-4"],"projectIds":["ur3-manipulation"],"tags":["ur3-manipulation","UR3 manipulation & vision labs","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"driver-monitoring-details","kind":"FACT","text":"Driver monitoring & diagnostic lab: Eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly detection; hierarchical face-loss and lens-occlusion fallbacks; speed, ignition, and gear gating; per-camera calibration; shadow-mode hard-case capture, human review, and replay-based metrics. Validated on real cabin video and shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics in lockstep with model development.","evidence":["benan-dms","driver-monitoring-catalog-1","driver-monitoring-catalog-2","driver-monitoring-catalog-3","driver-monitoring-catalog-4","driver-monitoring-catalog-5"],"projectIds":["driver-monitoring"],"tags":["driver-monitoring","Driver monitoring & diagnostic lab","implementation","technical depth","architecture","result","Python","MediaPipe","OpenCV","Temporal state machines","Calibration","Replay tooling"]},{"id":"driver-monitoring-limits","kind":"FACT","text":"Driver monitoring & diagnostic lab — evidence boundaries: Employment, on-device deployment, and cabin-video validation reflect Donald’s latest internship record. The 616-test suite is a project result, not a new website-run test. The supplied record does not claim safety certification, verified CAN integration, or quantified fleet-wide sensitivity/specificity.","evidence":["benan-dms","driver-monitoring-catalog-1","driver-monitoring-catalog-2","driver-monitoring-catalog-3","driver-monitoring-catalog-4","driver-monitoring-catalog-5"],"projectIds":["driver-monitoring"],"tags":["driver-monitoring","Driver monitoring & diagnostic lab","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"dms-cv181x-details","kind":"FACT","text":"DMS on CV181x: Sparse SCRFD detection with per-frame landmark tracking and ROI updates; legacy 478-point and 68-point backends; callbacks and display work moved off the per-frame critical path; timestamp-window PERCLOS; per-stage latency and tracking instrumentation; ONNX to CV181x BF16 conversion. Validated the end-to-end pipeline on real cabin video, with inference-cadence checks and per-stage compute budgeting. The internship record connects this edge pipeline to shadow-mode hard-case capture, review, and replay-based reporting.","evidence":["benan-dms","dms-cv181x-catalog-1","dms-cv181x-catalog-2","dms-cv181x-catalog-3","dms-cv181x-catalog-4"],"projectIds":["dms-cv181x"],"tags":["dms-cv181x","DMS on CV181x","implementation","technical depth","architecture","result","C","CV181x","SCRFD","PFLD","TPU-MLIR","ONNX","Embedded inference"]},{"id":"dms-cv181x-limits","kind":"FACT","text":"DMS on CV181x — evidence boundaries: The latest candidate-provided internship record updates the earlier repository-only hardware-validation status. It describes on-device deployment and end-to-end cabin-video validation; no numeric latency or accuracy benchmark is supplied. 20 ms remains a target, not a claimed achieved measurement. SCRFD and PFLD are upstream models; original model training is not claimed. Company-specific source is not mirrored on this website. The code computes sample ratios inside a timestamp window; do not describe it as exact time-integrated PERCLOS.","evidence":["benan-dms","dms-cv181x-catalog-1","dms-cv181x-catalog-2","dms-cv181x-catalog-3","dms-cv181x-catalog-4"],"projectIds":["dms-cv181x"],"tags":["dms-cv181x","DMS on CV181x","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"vggt-reconstruction-details","kind":"FACT","text":"VGGT scene reconstruction: ONNX sky segmentation, VGGT depth and camera prediction, depth unprojection, point filtering, and interactive inspection of the reconstructed geometry. Reconstruction and visualization scripts produce inspectable point-cloud artifacts with sky filtering.","evidence":["vggt-reconstruction-catalog-1","vggt-reconstruction-catalog-2","vggt-reconstruction-catalog-3"],"projectIds":["vggt-reconstruction"],"tags":["vggt-reconstruction","VGGT scene reconstruction","implementation","technical depth","architecture","result","PyTorch","VGGT","ONNX Runtime","3D geometry","Viser"]},{"id":"vggt-reconstruction-limits","kind":"FACT","text":"VGGT scene reconstruction — evidence boundaries: The original VGGT architecture and model weights are upstream. Local script presence does not establish a new model or a quantitative reconstruction result. Reconstruction quality and physical-scale accuracy remain unbenchmarked.","evidence":["vggt-reconstruction-catalog-1","vggt-reconstruction-catalog-2","vggt-reconstruction-catalog-3"],"projectIds":["vggt-reconstruction"],"tags":["vggt-reconstruction","VGGT scene reconstruction","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"sam3-experiments-details","kind":"FACT","text":"SAM 3 segmentation experiments: Text and geometric prompts, confidence thresholds, image-mask inspection, and CUDA mixed-precision inference in the supplied notebook. A locally adapted notebook for inspecting prompt-based segmentation, without a new training or benchmark result.","evidence":["sam3-experiments-catalog-1","sam3-experiments-catalog-2"],"projectIds":["sam3-experiments"],"tags":["sam3-experiments","SAM 3 segmentation experiments","implementation","technical depth","architecture","result","SAM 3","PyTorch","Image segmentation","CUDA","Jupyter"]},{"id":"sam3-experiments-limits","kind":"FACT","text":"SAM 3 segmentation experiments — evidence boundaries: This is an upstream-model experiment, not original SAM 3 development. Changes are currently uncommitted and no independent performance evaluation was inspected.","evidence":["sam3-experiments-catalog-1","sam3-experiments-catalog-2"],"projectIds":["sam3-experiments"],"tags":["sam3-experiments","SAM 3 segmentation experiments","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"chronos-knn-details","kind":"FACT","text":"Time-series forecasting & retrieval: Feature fitting must respect time. Input and target units must match. A ranking metric can improve while the position behavior becomes worse. The research framework connects retrieval, forecasting, and decision-rule evaluation. Ablations exposed an objective that disabled simulated exits, while regime-selected retraining did not outperform its rolling baseline in the documented study.","evidence":["chronos-architecture","chronos-failure","chronos-regime","chronos-knn-catalog-1","chronos-knn-catalog-2","chronos-knn-catalog-3","chronos-knn-catalog-4"],"projectIds":["chronos-knn"],"tags":["chronos-knn","Time-series forecasting & retrieval","implementation","technical depth","architecture","result","PyTorch","Qdrant","Time series","Walk-forward evaluation","Chronos","Kronos","Granite TTM","LoRA"]},{"id":"chronos-knn-limits","kind":"FACT","text":"Time-series forecasting & retrieval — evidence boundaries: Upstream foundation models are not Donald inventions. No live-performance claim. Architecture time filters alone do not certify full label-availability correctness; simulator leverage and calibration caveats preclude promoting headline returns. Later persistence heads are not proven to fix the earlier failure.","evidence":["chronos-architecture","chronos-failure","chronos-regime","chronos-knn-catalog-1","chronos-knn-catalog-2","chronos-knn-catalog-3","chronos-knn-catalog-4"],"projectIds":["chronos-knn"],"tags":["chronos-knn","Time-series forecasting & retrieval","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"momentum-transformer-adaptation-details","kind":"FACT","text":"Momentum Transformer & execution learning: Attention/LSTM financial sequence modeling, direct risk-objective optimization, regime features, portfolio-aware action projection, PPO/SAC training, GAE, observation normalization, transaction-cost accounting and execution adapters. Portfolio and execution-agent interfaces support threshold and reinforcement-learning policies. The work also corrected clipped-action log-probabilities and duplicate transaction penalties in the simulation/training path.","evidence":["momentum-transformer-adaptation-catalog-1","momentum-transformer-adaptation-catalog-2","momentum-transformer-adaptation-catalog-3","momentum-transformer-adaptation-catalog-4"],"projectIds":["momentum-transformer-adaptation"],"tags":["momentum-transformer-adaptation","Momentum Transformer & execution learning","implementation","technical depth","architecture","result","PyTorch","Temporal attention","LSTM","PPO","SAC","Execution simulation"]},{"id":"momentum-transformer-adaptation-limits","kind":"FACT","text":"Momentum Transformer & execution learning — evidence boundaries: Original papers and their reported results belong to upstream authors. Names such as production execution layer denote code organization, not verified deployed profitability. No model or trading tests rerun in this audit.","evidence":["momentum-transformer-adaptation-catalog-1","momentum-transformer-adaptation-catalog-2","momentum-transformer-adaptation-catalog-3","momentum-transformer-adaptation-catalog-4"],"projectIds":["momentum-transformer-adaptation"],"tags":["momentum-transformer-adaptation","Momentum Transformer & execution learning","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"mini-torch-details","kind":"FACT","text":"mini_torch — autograd from NumPy: Function/Context graph nodes; gradient accumulation along multiple paths; unbroadcasting; batched matmul derivatives; reductions and activation derivatives; numerically stabilized softmax cross-entropy; module/parameter discovery; dataset/batching; NumPy-only optimizer and train/test preprocessing. An end-to-end teaching library spanning differentiation, batching, optimization, and MNIST training code, with supplied unit tests.","evidence":["mini-torch-catalog-1","mini-torch-catalog-2","mini-torch-catalog-3","mini-torch-catalog-4","mini-torch-catalog-5"],"projectIds":["mini-torch"],"tags":["mini-torch","mini_torch — autograd from NumPy","implementation","technical depth","architecture","result","Python","NumPy","Automatic differentiation","Neural networks","MNIST"]},{"id":"mini-torch-limits","kind":"FACT","text":"mini_torch — autograd from NumPy — evidence boundaries: A CS446/ECE449 course study based on supplied scaffolding, not original autodiff research. No Git history or explicit candidate-named report was found; public phrasing should retain the coursework attribution. Do not claim generic PyTorch compatibility or production performance. No measured classification accuracy is claimed.","evidence":["mini-torch-catalog-1","mini-torch-catalog-2","mini-torch-catalog-3","mini-torch-catalog-4","mini-torch-catalog-5"],"projectIds":["mini-torch"],"tags":["mini-torch","mini_torch — autograd from NumPy","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"rl-portfolio-framework-details","kind":"FACT","text":"Regime-aware reinforcement learning: Recurrent policies, PID-Lagrangian drawdown constraints, holdings/turnover-aware simulation, online versus oracle regimes, event-gated entry, reporting-lag alignment, fundamental transforms and controlled signal ablations. Portfolio environments combine PPO-PID drawdown constraints, turnover-aware simulation, regime modules, and diagnostics. The research explicitly separates causal online labels from smoothed oracle labels.","evidence":["rl-portfolio-framework-catalog-1","rl-portfolio-framework-catalog-2","rl-portfolio-framework-catalog-3","rl-portfolio-framework-catalog-4","rl-portfolio-framework-catalog-5","rl-portfolio-framework-catalog-6","rl-portfolio-framework-catalog-7","rl-portfolio-framework-catalog-8"],"projectIds":["rl-portfolio-framework"],"tags":["rl-portfolio-framework","Regime-aware reinforcement learning","implementation","technical depth","architecture","result","PyTorch","Gymnasium","PPO","RecurrentPPO","Jump models","GMM","Streamlit","Python","Tushare","Logistic regression","Fundamental factors"]},{"id":"rl-portfolio-framework-limits","kind":"FACT","text":"Regime-aware reinforcement learning — evidence boundaries: No production-readiness, profitability or live-computability certification. Reference-paper presence is not authorship or completed reproduction. Theoretical prototypes and deployed execution must remain separate.","evidence":["rl-portfolio-framework-catalog-1","rl-portfolio-framework-catalog-2","rl-portfolio-framework-catalog-3","rl-portfolio-framework-catalog-4","rl-portfolio-framework-catalog-5","rl-portfolio-framework-catalog-6","rl-portfolio-framework-catalog-7","rl-portfolio-framework-catalog-8"],"projectIds":["rl-portfolio-framework"],"tags":["rl-portfolio-framework","Regime-aware reinforcement learning","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"xtrend-details","kind":"FACT","text":"X-Trend reproduction: A tensor with the right shape can encode the wrong meaning. Paper alignment requires checking both data availability and the information flowing through attention. The revised PyTorch construction lets retrieved context carry both market conditions and their observed outcomes. Experiment notes also record numerical instability in change-point fitting and later segmentation changes; original-paper performance remains unverified.","evidence":["xtrend-values","xtrend-source","xtrend-catalog-1","xtrend-catalog-2","xtrend-catalog-3","xtrend-catalog-4","xtrend-catalog-5"],"projectIds":["xtrend"],"tags":["xtrend","X-Trend reproduction","implementation","technical depth","architecture","result","PyTorch","Cross-attention","LSTM","Expanding windows","Gaussian processes","Sparse jump models"]},{"id":"xtrend-limits","kind":"FACT","text":"X-Trend reproduction — evidence boundaries: Do not repeat the early README claim of superior returns or commit subjects naming target Sharpe as achieved performance. Architecture reproduction, economic reproduction and live trading remain distinct.","evidence":["xtrend-values","xtrend-source","xtrend-catalog-1","xtrend-catalog-2","xtrend-catalog-3","xtrend-catalog-4","xtrend-catalog-5"],"projectIds":["xtrend"],"tags":["xtrend","X-Trend reproduction","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"generative-models-details","kind":"FACT","text":"VAE, DDPM & guided diffusion: Convolutional VAE with KL weighting and generation-distribution comparison; time-conditioned U-Net diffusion from supplied scaffolding; two-pass classifier-free guidance; Tweedie reconstruction; self-adjoint blur likelihood correction; six guidance scales and four noise settings. Generated-image artifacts and a qualitative deblurring study show the tradeoff: insufficient guidance loses the input, excessive guidance introduces artifacts, and higher measurement noise degrades reconstruction.","evidence":["generative-models-catalog-1","generative-models-catalog-2","generative-models-catalog-3","generative-models-catalog-4","generative-models-catalog-5","generative-models-catalog-6"],"projectIds":["generative-models"],"tags":["generative-models","VAE, DDPM & guided diffusion","implementation","technical depth","architecture","result","Python","PyTorch","VAE","DDPM","Classifier-free guidance","Diffusion posterior sampling","Jupyter"]},{"id":"generative-models-limits","kind":"FACT","text":"VAE, DDPM & guided diffusion — evidence boundaries: Coursework using supplied architectures and OpenAI guided-diffusion/pretrained weights; those upstream models are not Donald-authored. The DPS report is qualitative and based on its selected image setup, not broad image-restoration performance. No aggregate PSNR/SSIM or reproducibility benchmark was established. mp2 lowercase is a setup duplicate; group it with MP2 rather than count it as another project.","evidence":["generative-models-catalog-1","generative-models-catalog-2","generative-models-catalog-3","generative-models-catalog-4","generative-models-catalog-5","generative-models-catalog-6"],"projectIds":["generative-models"],"tags":["generative-models","VAE, DDPM & guided diffusion","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"adaptive-portfolio-learning-details","kind":"FACT","text":"A-share adaptive portfolio learning: Causal support/query adaptation, monthly retraining, market-specific execution constraints, utility labels, feasible teachers, DAgger scaffolding, conformal lower bounds, reliability abstention and walk-forward tests. Modular pipelines support walk-forward portfolio experiments. The documented DeePM study rejected the tested long-only selection thesis and separated market exposure from active selection value.","evidence":["adaptive-portfolio-learning-catalog-1","adaptive-portfolio-learning-catalog-2","adaptive-portfolio-learning-catalog-3","adaptive-portfolio-learning-catalog-4","adaptive-portfolio-learning-catalog-5"],"projectIds":["adaptive-portfolio-learning"],"tags":["adaptive-portfolio-learning","A-share adaptive portfolio learning","implementation","technical depth","architecture","result","PyTorch","DoubleAdapt","StockMixer","DeePM","Imitation learning","Conformal calibration","LightGBM"]},{"id":"adaptive-portfolio-learning-limits","kind":"FACT","text":"A-share adaptive portfolio learning — evidence boundaries: Original models and publications remain upstream. The negative DeePM report has universe/survivorship caveats and should be presented as its dated experiment, not a universal impossibility result. No headline financial metrics are promoted.","evidence":["adaptive-portfolio-learning-catalog-1","adaptive-portfolio-learning-catalog-2","adaptive-portfolio-learning-catalog-3","adaptive-portfolio-learning-catalog-4","adaptive-portfolio-learning-catalog-5"],"projectIds":["adaptive-portfolio-learning"],"tags":["adaptive-portfolio-learning","A-share adaptive portfolio learning","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"project-genji-details","kind":"FACT","text":"Project Genji: temporal model evaluation: Cross-sectional winsorization, industry normalization, multi-horizon training, purged/embargoed temporal splits, deterministic model comparison and structured prediction/evaluation contracts. Ridge and XGBoost training share purged and embargoed temporal splits, fold metrics, and aggregated out-of-fold predictions for structured model comparison.","evidence":["project-genji-catalog-1","project-genji-catalog-2","project-genji-catalog-3","project-genji-catalog-4"],"projectIds":["project-genji"],"tags":["project-genji","Project Genji: temporal model evaluation","implementation","technical depth","architecture","result","Python","Qlib","XGBoost","Ridge regression","Purged cross-validation","Tushare"]},{"id":"project-genji-limits","kind":"FACT","text":"Project Genji: temporal model evaluation — evidence boundaries: Do not call all originally planned regime, QP portfolio, RL and backtest components complete. README plans and current implemented supervised-learning path must be distinguished.","evidence":["project-genji-catalog-1","project-genji-catalog-2","project-genji-catalog-3","project-genji-catalog-4"],"projectIds":["project-genji"],"tags":["project-genji","Project Genji: temporal model evaluation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"symbolic-alpha-mining-details","kind":"FACT","text":"Symbolic alpha mining and allocation: Symbolic expression trees, alpha pools, knowledge-guided search, GFlowNet/PPO search entrypoints, point-in-time feature access and allocation after discovery. The workspace connects symbolic-alpha search to explicit feature-availability controls and allocation code, extending the upstream discovery workflow.","evidence":["symbolic-alpha-mining-catalog-1","symbolic-alpha-mining-catalog-2","symbolic-alpha-mining-catalog-3"],"projectIds":["symbolic-alpha-mining"],"tags":["symbolic-alpha-mining","Symbolic alpha mining and allocation","implementation","technical depth","architecture","result","Python","Symbolic expressions","GFlowNet","PPO","Point-in-time data"]},{"id":"symbolic-alpha-mining-limits","kind":"FACT","text":"Symbolic alpha mining and allocation — evidence boundaries: Original AlphaPROBE paper/model is not Donald’s publication. Working-tree additions establish local work, not validated alpha or exact sole authorship. No research outputs inspected.","evidence":["symbolic-alpha-mining-catalog-1","symbolic-alpha-mining-catalog-2","symbolic-alpha-mining-catalog-3"],"projectIds":["symbolic-alpha-mining"],"tags":["symbolic-alpha-mining","Symbolic alpha mining and allocation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"visual-alpha-research-details","kind":"FACT","text":"A-share representation and attention-model adaptations: Reproducible chart rendering, visual-versus-numeric comparisons, market-conditioned adapters, characteristic attention, variable-universe masks, PIT alignment and portfolio construction under market constraints. The family includes chart rendering, model and evaluation modules, and ViT inference/visualization. Attention Factors distinguishes the paper’s long/short reference from a long-tilt hedge construction.","evidence":["visual-alpha-research-catalog-1","visual-alpha-research-catalog-2","visual-alpha-research-catalog-3","visual-alpha-research-catalog-4","visual-alpha-research-catalog-5","visual-alpha-research-catalog-6","visual-alpha-research-catalog-7"],"projectIds":["visual-alpha-research"],"tags":["visual-alpha-research","A-share representation and attention-model adaptations","implementation","technical depth","architecture","result","PyTorch","CNN","Vision Transformer","MASTER","t-SNE","A-share data","Attention factors","Statistical arbitrage","Point-in-time panels"]},{"id":"visual-alpha-research-limits","kind":"FACT","text":"A-share representation and attention-model adaptations — evidence boundaries: No paper publication, successful SDF replication, model advantage or deployment is established. Planned visual fusion and directly executable market variants must not be conflated with finished evaluation.","evidence":["visual-alpha-research-catalog-1","visual-alpha-research-catalog-2","visual-alpha-research-catalog-3","visual-alpha-research-catalog-4","visual-alpha-research-catalog-5","visual-alpha-research-catalog-6","visual-alpha-research-catalog-7"],"projectIds":["visual-alpha-research"],"tags":["visual-alpha-research","A-share representation and attention-model adaptations","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"rotation-learning-details","kind":"FACT","text":"Leader-follower rotation and event ranking: Within-event ranking, matured-label graph updates, temporal encoders, filtered HMM posteriors, weighted Plackett-Luce choice, out-of-fold calibration, lower-confidence-bound expected value and abstention. Alternative ranking paths use matured labels, filtered states, and explicit feature-availability rules. The GNN path excludes present-day industry classifications because they are not historical point-in-time labels.","evidence":["rotation-learning-catalog-1","rotation-learning-catalog-2","rotation-learning-catalog-3","rotation-learning-catalog-4","rotation-learning-catalog-5"],"projectIds":["rotation-learning"],"tags":["rotation-learning","Leader-follower rotation and event ranking","implementation","technical depth","architecture","result","GNN","Student-t HMM","Temporal encoders","Plackett-Luce","Calibration","Event ranking"]},{"id":"rotation-learning-limits","kind":"FACT","text":"Leader-follower rotation and event ranking — evidence boundaries: No leaderboard or trading-performance claims. Production wording in local model docs does not establish deployment. Do not combine experiment variants into a single validated algorithm.","evidence":["rotation-learning-catalog-1","rotation-learning-catalog-2","rotation-learning-catalog-3","rotation-learning-catalog-4","rotation-learning-catalog-5"],"projectIds":["rotation-learning"],"tags":["rotation-learning","Leader-follower rotation and event ranking","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"cnn-ablation-harness-details","kind":"FACT","text":"CIFAR-10 experiment harness: Reproducibility controls; configurable affine transforms; architecture-width variants; train/evaluation loop; machine-readable result output; modular ablation harness. A configurable CIFAR-10 experiment harness with separated training/evaluation modules, machine-readable results, and smoke-test source. README sample scores are illustrative rather than measured outcomes.","evidence":["cnn-ablation-harness-catalog-1","cnn-ablation-harness-catalog-2","cnn-ablation-harness-catalog-3","cnn-ablation-harness-catalog-4","cnn-ablation-harness-catalog-5"],"projectIds":["cnn-ablation-harness"],"tags":["cnn-ablation-harness","CIFAR-10 experiment harness","implementation","technical depth","architecture","result","Python","PyTorch","CIFAR-10","Experiment configuration"]},{"id":"cnn-ablation-harness-limits","kind":"FACT","text":"CIFAR-10 experiment harness — evidence boundaries: Tutorial-based course work, not a new CNN architecture. No independent candidate attribution or fresh test run. The course variation applies augmentation to both training and test transforms, so comparisons need that evaluation caveat.","evidence":["cnn-ablation-harness-catalog-1","cnn-ablation-harness-catalog-2","cnn-ablation-harness-catalog-3","cnn-ablation-harness-catalog-4","cnn-ablation-harness-catalog-5"],"projectIds":["cnn-ablation-harness"],"tags":["cnn-ablation-harness","CIFAR-10 experiment harness","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"upstream-ai-reference-library-details","kind":"FACT","text":"Forecasting & market-agent reference library: Reference topics are multimodal/counterfactual forecasting and LLM-based market-agent systems; these describe upstream scope only. Reference implementations for multimodal forecasting and LLM market agents, cataloged separately from the local World Cup announcer-mention study.","evidence":["upstream-ai-reference-library-catalog-1","upstream-ai-reference-library-catalog-2","upstream-ai-reference-library-catalog-3","upstream-ai-reference-library-catalog-4","upstream-ai-reference-library-catalog-5"],"projectIds":["upstream-ai-reference-library"],"tags":["upstream-ai-reference-library","Forecasting & market-agent reference library","implementation","technical depth","architecture","result","Multimodal learning","Causal forecasting","Counterfactual augmentation","Python","LLM applications","Prediction markets"]},{"id":"upstream-ai-reference-library-limits","kind":"FACT","text":"Forecasting & market-agent reference library — evidence boundaries: Neither upstream publication, KDD acceptance, original model, codebase, release, user count nor performance is Donald’s. Folder presence alone does not establish completed study or endorsement.","evidence":["upstream-ai-reference-library-catalog-1","upstream-ai-reference-library-catalog-2","upstream-ai-reference-library-catalog-3","upstream-ai-reference-library-catalog-4","upstream-ai-reference-library-catalog-5"],"projectIds":["upstream-ai-reference-library"],"tags":["upstream-ai-reference-library","Forecasting & market-agent reference library","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"local-qwen-inference-details","kind":"FACT","text":"Local Qwen inference: Durable call reservations use file locks and fsync, including failed requests. Decoding parameters are pinned, model/runtime metadata is captured, JSON-schema output is requested, and completion budgets scale with target count. The client explicitly disables reasoning for concise structured outputs after discovering null-content/truncation behavior. Local Qwen generation was exercised successfully. The client handles output truncation and distinguishes transport, HTTP, parsing, and completion failures; this is inference integration, not a new model or serving engine.","evidence":["local-qwen-inference-catalog-1","local-qwen-inference-catalog-2","local-qwen-inference-catalog-3","local-qwen-inference-catalog-4"],"projectIds":["local-qwen-inference"],"tags":["local-qwen-inference","Local Qwen inference","implementation","technical depth","architecture","result","Qwen","vLLM","OpenAI-compatible HTTP","JSON Schema","Python","fcntl","fsync","Quantization","PyTorch"]},{"id":"local-qwen-inference-limits","kind":"FACT","text":"Local Qwen inference — evidence boundaries: This is a capability collection across projects, not a separate published model or original inference engine. No throughput benchmark, SLA, model-weight provenance audit, external-customer deployment, or training claim is established. The autoresearch client is present in the working tree but has no path-specific committed attribution in the inspected history.","evidence":["local-qwen-inference-catalog-1","local-qwen-inference-catalog-2","local-qwen-inference-catalog-3","local-qwen-inference-catalog-4"],"projectIds":["local-qwen-inference"],"tags":["local-qwen-inference","Local Qwen inference","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"zhengmind-yagni-details","kind":"FACT","text":"ZhengMindYAGNI — agent orchestration: Dependency-ready claims use O_EXCL file locks. The core separates work-tree structure, transition rules, events, and persistence. The web conductor invokes planner/worker/reviewer processes while the CLI remains the state-mutation boundary; lesson memory and Git-head/live-deploy comparison make unfinished work visible. A dependency-free Go coordinator for planner, worker, and reviewer processes, with durable work nodes and transition/claim tests. Work records show the workflow applied to the YouTube Knowledge project.","evidence":["zhengmind-yagni-catalog-1","zhengmind-yagni-catalog-2","zhengmind-yagni-catalog-3","zhengmind-yagni-catalog-4","zhengmind-yagni-catalog-5","zhengmind-yagni-catalog-6"],"projectIds":["zhengmind-yagni"],"tags":["zhengmind-yagni","ZhengMindYAGNI — agent orchestration","implementation","technical depth","architecture","result","Go","Git","Markdown","JSONL","File locks","CLI","HTTP","Claude CLI"]},{"id":"zhengmind-yagni-limits","kind":"FACT","text":"ZhengMindYAGNI — agent orchestration — evidence boundaries: README statements of about 1.6k lines and a purely read-only dashboard are stale: current source has more code and POST endpoints that start agent loops. No production, multi-tenant, distributed coordination, or external-adoption claim is established. The architectural critique is project documentation, not a verified first-person quotation. The coordination engine is local; diagnostic agent calls use Claude CLI and are not local-model inference.","evidence":["zhengmind-yagni-catalog-1","zhengmind-yagni-catalog-2","zhengmind-yagni-catalog-3","zhengmind-yagni-catalog-4","zhengmind-yagni-catalog-5","zhengmind-yagni-catalog-6"],"projectIds":["zhengmind-yagni"],"tags":["zhengmind-yagni","ZhengMindYAGNI — agent orchestration","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"vta-qwen-adaptation-details","kind":"FACT","text":"VTA — local Qwen adaptation: Typed chat content and processor normalization handle Qwen3.5 multimodal architecture conventions. Loader/generation paths distinguish FastModel from FastLanguageModel and disable the vLLM fast path for this architecture. Pipeline stages combine reward training, rejection sampling, supervised fine-tuning, and evaluation. Compatibility code spans generation and staged GRPO/LoRA/SFT workflows. A completed training run or forecasting improvement has not been established.","evidence":["vta-qwen-adaptation-catalog-1","vta-qwen-adaptation-catalog-2","vta-qwen-adaptation-catalog-3","vta-qwen-adaptation-catalog-4","vta-qwen-adaptation-catalog-5"],"projectIds":["vta-qwen-adaptation"],"tags":["vta-qwen-adaptation","VTA — local Qwen adaptation","implementation","technical depth","architecture","result","Qwen3.5","Unsloth","PyTorch","Transformers","TRL","LoRA","GRPO","SFT","CUDA","Python"]},{"id":"vta-qwen-adaptation-limits","kind":"FACT","text":"VTA — local Qwen adaptation — evidence boundaries: Upstream algorithm, training pipeline, and paper authorship must be credited separately. README marketing about improved forecasting is not a verified result of this adaptation. Local compatibility branches are environment-specific and not a universal statement that Qwen3.5 cannot run on vLLM. No predictive advantage, live trading value, or training completion is established.","evidence":["vta-qwen-adaptation-catalog-1","vta-qwen-adaptation-catalog-2","vta-qwen-adaptation-catalog-3","vta-qwen-adaptation-catalog-4","vta-qwen-adaptation-catalog-5"],"projectIds":["vta-qwen-adaptation"],"tags":["vta-qwen-adaptation","VTA — local Qwen adaptation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"qwen-market-ending-details","kind":"FACT","text":"Event-ending inference with Qwen: The pipeline groups related markets around a real-world event, hides the target moneyline, separates official feed facts from market-price evidence, and supports monotonic multi-horizon outputs. Uncertainty is an explicit downstream risk state rather than forced certainty. A dated 25-game exploratory comparison reported 22 judged-correct decisions and three unsure. The small sample and market-close labels do not establish general accuracy or economic value.","evidence":["qwen-market-ending-catalog-1","qwen-market-ending-catalog-2","qwen-market-ending-catalog-3","qwen-market-ending-catalog-4"],"projectIds":["qwen-market-ending"],"tags":["qwen-market-ending","Event-ending inference with Qwen","implementation","technical depth","architecture","result","Python standard library","Qwen","vLLM","HTTP APIs","JSON","Concurrent futures","Evaluation"]},{"id":"qwen-market-ending-limits","kind":"FACT","text":"Event-ending inference with Qwen — evidence boundaries: Small opportunistic sample with mostly straightforward positives; no robust generalization or economic-value claim. The documented ground truth uses market close time, which should not be assumed identical to real event end without further validation. Cross-exchange game identity is not fully unified.","evidence":["qwen-market-ending-catalog-1","qwen-market-ending-catalog-2","qwen-market-ending-catalog-3","qwen-market-ending-catalog-4"],"projectIds":["qwen-market-ending"],"tags":["qwen-market-ending","Event-ending inference with Qwen","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"semantic-research-agents-details","kind":"FACT","text":"Semantic research agents: Structured model outputs, point-in-time evidence, research orchestration, confirmation versus veto logic, and constrained portfolio construction. A set of specialized research agents, structured decision modules, and dashboards. Confirmation signals, vetoes, and portfolio constraints make the proposed decisions inspectable; the experiments retain separate validation boundaries.","evidence":["semantic-research-agents-catalog-1","semantic-research-agents-catalog-2","semantic-research-agents-catalog-3","semantic-research-agents-catalog-4"],"projectIds":["semantic-research-agents"],"tags":["semantic-research-agents","Semantic research agents","implementation","technical depth","architecture","result","Python","LLM agents","Structured outputs","TradingAgents","ATLAS","Tushare"]},{"id":"semantic-research-agents-limits","kind":"FACT","text":"Semantic research agents — evidence boundaries: Original frameworks and papers remain upstream. ATLAS subscriber/revenue claims and VTA state-of-the-art claims must never be attributed to Donald. LLM reasoning quality, training success, financial performance and adoption are unverified.","evidence":["semantic-research-agents-catalog-1","semantic-research-agents-catalog-2","semantic-research-agents-catalog-3","semantic-research-agents-catalog-4"],"projectIds":["semantic-research-agents"],"tags":["semantic-research-agents","Semantic research agents","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"agent-context-blueprint-details","kind":"FACT","text":"Typed handoffs for coding agents: Patch packages carry tickets, diffs, tests, and review context. Hooks guard direct writes, build context packs, enforce branch/prompt policies, invoke Codex review, and drive a test/integration pipeline. Context windows are separated by role and token budgets are explicit. Patch packages carry tickets, diffs, tests, and review context. Deterministic hooks construct context packs, enforce workflow boundaries, and drive review and integration.","evidence":["agent-context-blueprint-catalog-1","agent-context-blueprint-catalog-2","agent-context-blueprint-catalog-3","agent-context-blueprint-catalog-4"],"projectIds":["agent-context-blueprint"],"tags":["agent-context-blueprint","Typed handoffs for coding agents","implementation","technical depth","architecture","result","Python","Shell","JSON Schema","Git","Claude Code hooks","Codex","Agent orchestration"]},{"id":"agent-context-blueprint-limits","kind":"FACT","text":"Typed handoffs for coding agents — evidence boundaries: A workflow/template is not a hosted product or evidence of autonomous reliability at scale. Some starter assets and examples are scaffolding, not deployed application work. Hooks execute commands and require project-specific review; no runtime or security validation was performed.","evidence":["agent-context-blueprint-catalog-1","agent-context-blueprint-catalog-2","agent-context-blueprint-catalog-3","agent-context-blueprint-catalog-4"],"projectIds":["agent-context-blueprint"],"tags":["agent-context-blueprint","Typed handoffs for coding agents","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"claude-code-autopilot-details","kind":"FACT","text":"Claude Code Autopilot: Prompt hooks select relevant skills, session hooks restore context, installation backs up existing files and installs hook dependencies, and review workflow was revised away from a blocking hook toward a skill-based approach. A reusable setup script backs up existing files, installs hook dependencies, and activates context and review workflows for coding projects.","evidence":["claude-code-autopilot-catalog-1","claude-code-autopilot-catalog-2","claude-code-autopilot-catalog-3","claude-code-autopilot-catalog-4"],"projectIds":["claude-code-autopilot"],"tags":["claude-code-autopilot","Claude Code Autopilot","implementation","technical depth","architecture","result","Shell","Python","TypeScript","Claude Code","Hooks","Codex","Skills"]},{"id":"claude-code-autopilot-limits","kind":"FACT","text":"Claude Code Autopilot — evidence boundaries: Upstream skill counts and production-tested wording belong to the integrated ecosystems, not independent proof of Donald deployment outcomes. No quantified productivity benefit or broad installation compatibility was established.","evidence":["claude-code-autopilot-catalog-1","claude-code-autopilot-catalog-2","claude-code-autopilot-catalog-3","claude-code-autopilot-catalog-4"],"projectIds":["claude-code-autopilot"],"tags":["claude-code-autopilot","Claude Code Autopilot","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"zhengmind-foundation-details","kind":"FACT","text":"ZhengMind architecture study: Task trees, dependency-aware claims, scoped API tokens, prompt snapshots, sanitized audit events, proposal/SCM seams, and explicit parent integration express a sophisticated agent-governance model. The reference prototype explores task trees, dependency-aware claims, prompt snapshots, and auditable state changes. Its external database, GitHub-write, and LLM adapters remain unconnected.","evidence":["zhengmind-foundation-catalog-1","zhengmind-foundation-catalog-2","zhengmind-foundation-catalog-3","zhengmind-foundation-catalog-4","zhengmind-foundation-catalog-5"],"projectIds":["zhengmind-foundation"],"tags":["zhengmind-foundation","ZhengMind architecture study","implementation","technical depth","architecture","result","Python","FastAPI","SQLite","Jinja2","Task graphs","Scoped tokens","Audit trails"]},{"id":"zhengmind-foundation-limits","kind":"FACT","text":"ZhengMind architecture study — evidence boundaries: Not a production deployment; real external adapters and LLM execution are not established. Some domain concepts are future plans rather than implemented runtime support. Historical test counts in README were not rerun.","evidence":["zhengmind-foundation-catalog-1","zhengmind-foundation-catalog-2","zhengmind-foundation-catalog-3","zhengmind-foundation-catalog-4","zhengmind-foundation-catalog-5"],"projectIds":["zhengmind-foundation"],"tags":["zhengmind-foundation","ZhengMind architecture study","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"serena-foundation-details","kind":"FACT","text":"Serena semantic tooling reference: Upstream Serena connects language-server symbol information to agent tools such as find_symbol, find_referencing_symbols, and insert_after_symbol; MCP decouples the tools from a particular model/client. A concrete reference for connecting symbol definitions and references to model-facing editing tools.","evidence":["serena-foundation-catalog-1","serena-foundation-catalog-2"],"projectIds":["serena-foundation"],"tags":["serena-foundation","Serena semantic tooling reference","implementation","technical depth","architecture","result","Python","MCP","Language Server Protocol","Semantic code tools"]},{"id":"serena-foundation-limits","kind":"FACT","text":"Serena semantic tooling reference — evidence boundaries: Workspace presence does not prove extensive use, deployment, or original work. All capability, adoption, benchmark, and productivity claims belong to upstream unless separately substantiated.","evidence":["serena-foundation-catalog-1","serena-foundation-catalog-2"],"projectIds":["serena-foundation"],"tags":["serena-foundation","Serena semantic tooling reference","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"isafe-sop-details","kind":"FACT","text":"ISAFE / WITNESS — SOP verification: SOP steps form a DAG; sustained observations debounce noisy detections. Mandatory-ancestor precedence distinguishes alternate paths, skipped steps, and late wrong-order completions. Soft-DTW is downsampled and normalized to bound per-frame work. Evidence clips feed a second-opinion adjudicator, with a hash-chained audit ledger and SQLite/Postgres seams. The prototype connects few-shot registration, step conformance, video evidence, adjudication, persistence, and a review console. Its documented mock-perception acceptance scenario deliberately skips assembly step C to exercise procedure checking.","evidence":["isafe-sop-catalog-1","isafe-sop-catalog-2","isafe-sop-catalog-3","isafe-sop-catalog-4","isafe-sop-catalog-5","isafe-sop-catalog-6"],"projectIds":["isafe-sop"],"tags":["isafe-sop","ISAFE / WITNESS — SOP verification","implementation","technical depth","architecture","result","Python","FastAPI","Pydantic","NumPy","Soft-DTW","DAG/FSM","Qwen2.5-VL","PyTorch","bitsandbytes","DINOv2","FAISS","SQLite","PostgreSQL","MQTT","WebSocket"]},{"id":"isafe-sop-limits","kind":"FACT","text":"ISAFE / WITNESS — SOP verification — evidence boundaries: Mock CPU backends are the default. Real GPU inference and industrial deployment were not demonstrated by this audit. The current natural-language SOP compiler is a deterministic keyword parser; its LLM-backed compiler is a TODO seam despite broader product-language claims. Qwen2.5-VL 3B uses an opt-in NF4 4-bit adapter with visual-token bounds; parser tests do not establish real model accuracy. Field-signal adapters are stubs; production safety or defect-reduction outcomes are not established.","evidence":["isafe-sop-catalog-1","isafe-sop-catalog-2","isafe-sop-catalog-3","isafe-sop-catalog-4","isafe-sop-catalog-5","isafe-sop-catalog-6"],"projectIds":["isafe-sop"],"tags":["isafe-sop","ISAFE / WITNESS — SOP verification","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"cpp-goroutine-details","kind":"FACT","text":"A Go-style runtime in C++: ucontext/makecontext/swapcontext scheduling across worker threads; mmap/mprotect stack guards; epoll and eventfd wakeups; deadline min-heap; read/write/recv/send/connect/accept/sleep hooks resolved with dlsym; user O_NONBLOCK tracking; parking handshake to avoid resuming before a context is saved; yielding mutexes and wait groups. A library with CMake targets and examples for scheduling, timers, and a loopback echo server.","evidence":["cpp-goroutine-catalog-1","cpp-goroutine-catalog-2","cpp-goroutine-catalog-3","cpp-goroutine-catalog-4","cpp-goroutine-catalog-5","cpp-goroutine-catalog-6","cpp-goroutine-catalog-7"],"projectIds":["cpp-goroutine"],"tags":["cpp-goroutine","A Go-style runtime in C++","implementation","technical depth","architecture","result","C++17","Linux","ucontext","epoll","eventfd","pthread","CMake"]},{"id":"cpp-goroutine-limits","kind":"FACT","text":"A Go-style runtime in C++ — evidence boundaries: Experimental Linux-specific runtime, not a Go implementation or production concurrency guarantee. Examples do not establish scheduler race-freedom, cancellation correctness, stress reliability or benchmark throughput. No independent authorship/date history found. Throughput and latency have not been benchmarked.","evidence":["cpp-goroutine-catalog-1","cpp-goroutine-catalog-2","cpp-goroutine-catalog-3","cpp-goroutine-catalog-4","cpp-goroutine-catalog-5","cpp-goroutine-catalog-6","cpp-goroutine-catalog-7"],"projectIds":["cpp-goroutine"],"tags":["cpp-goroutine","A Go-style runtime in C++","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"applypilot-adaptation-details","kind":"FACT","text":"ApplyPilot workflow adaptation: The CLI adapter passes prompts through stdin, selects a model, avoids session persistence, isolates settings/CWD, and retries failures. Scoring uses ThreadPoolExecutor and periodically commits completed batches so a later crash loses less finished work. Completed scoring batches are committed incrementally so a later failure loses less finished work. The model adapter avoids persistent sessions and isolates its settings and working directory.","evidence":["applypilot-adaptation-catalog-1","applypilot-adaptation-catalog-2","applypilot-adaptation-catalog-3","applypilot-adaptation-catalog-4"],"projectIds":["applypilot-adaptation"],"tags":["applypilot-adaptation","ApplyPilot workflow adaptation","implementation","technical depth","architecture","result","Python","httpx","Claude CLI","ThreadPoolExecutor","SQLite","Subprocess"]},{"id":"applypilot-adaptation-limits","kind":"FACT","text":"ApplyPilot workflow adaptation — evidence boundaries: No candidate-authored committed provenance or executed test result for these edits was established. CLI invocation is a cloud-model integration, not local inference; upstream local-provider support is not a new candidate invention. README application counts are upstream promotional claims and must not be attributed to Donald. No application records, résumé artifacts, scraped job records, or candidate profile data were inspected.","evidence":["applypilot-adaptation-catalog-1","applypilot-adaptation-catalog-2","applypilot-adaptation-catalog-3","applypilot-adaptation-catalog-4"],"projectIds":["applypilot-adaptation"],"tags":["applypilot-adaptation","ApplyPilot workflow adaptation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"invoice-authentication-details","kind":"FACT","text":"Full-stack authentication study: Prisma-backed refresh records; rotation transactions; HttpOnly cookie transport; GraphQL context/resolver integration; Inversify service boundaries; an alternative repository/token-service design; tests for cookie rotation and preventing refresh tokens in JSON. Authentication implementations span browser cookies, GraphQL context, and persistence, with targeted test source for rotation and keeping refresh tokens out of JSON responses.","evidence":["invoice-authentication-catalog-1","invoice-authentication-catalog-2","invoice-authentication-catalog-3","invoice-authentication-catalog-4","invoice-authentication-catalog-5"],"projectIds":["invoice-authentication"],"tags":["invoice-authentication","Full-stack authentication study","implementation","technical depth","architecture","result","TypeScript","React","Express","GraphQL","Prisma","PostgreSQL","JWT","Vitest"]},{"id":"invoice-authentication-limits","kind":"FACT","text":"Full-stack authentication study — evidence boundaries: Do not present the entire upstream invoice app, its original personal project narrative or CI/CD history as Donald’s work. Variants differ; do not combine them into a claim that one final implementation contains all safeguards. Security correctness and concurrency behavior were not independently validated. Skip timesheets, credentials, personal evaluation metadata and real invoice data. Production security is not established.","evidence":["invoice-authentication-catalog-1","invoice-authentication-catalog-2","invoice-authentication-catalog-3","invoice-authentication-catalog-4","invoice-authentication-catalog-5"],"projectIds":["invoice-authentication"],"tags":["invoice-authentication","Full-stack authentication study","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"cuda-introduction-details","kind":"FACT","text":"CUDA development environment study: CUDA compiler/toolchain, GPU device properties and batch scheduling exposure, as an introductory reference rather than a custom parallel-computing project. A concrete foundation for inspecting GPU device properties and understanding the CUDA build and batch-execution workflow.","evidence":["cuda-introduction-catalog-1","cuda-introduction-catalog-2","cuda-introduction-catalog-3"],"projectIds":["cuda-introduction"],"tags":["cuda-introduction","CUDA development environment study","implementation","technical depth","architecture","result","CUDA","C++","Slurm"]},{"id":"cuda-introduction-limits","kind":"FACT","text":"CUDA development environment study — evidence boundaries: Do not claim custom CUDA kernels, GPU optimization, performance results or original authorship from this folder. Represent as current learning/reference only, if included publicly.","evidence":["cuda-introduction-catalog-1","cuda-introduction-catalog-2","cuda-introduction-catalog-3"],"projectIds":["cuda-introduction"],"tags":["cuda-introduction","CUDA development environment study","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"blockchain-principles-details","kind":"FACT","text":"Blockchain principles reading collection: Reference material for distributed ledger principles; specific mastered topics or original systems are not established by file possession. A collected set of lecture material on distributed-ledger principles; no authored system or research result is claimed.","evidence":["blockchain-principles-catalog-1","blockchain-principles-catalog-2"],"projectIds":["blockchain-principles"],"tags":["blockchain-principles","Blockchain principles reading collection","implementation","technical depth","architecture","result","Distributed systems","Blockchain coursework"]},{"id":"blockchain-principles-limits","kind":"FACT","text":"Blockchain principles reading collection — evidence boundaries: Reference ownership is not original authorship, completion or demonstrated expertise. Files were inventoried, not deeply read or redistributed.","evidence":["blockchain-principles-catalog-1","blockchain-principles-catalog-2"],"projectIds":["blockchain-principles"],"tags":["blockchain-principles","Blockchain principles reading collection","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"kalshibot-details","kind":"FACT","text":"Prediction-market execution systems: Partial fills, unknown order intents, restarts, and cancellation ownership all have to preserve quantity-level exposure. An independently packaged execution/research system with reconciliation for ambiguous order outcomes. Dated September 4 verification records 1,526 passing offline tests; production equivalence and profitability are separate, unproven outcomes.","evidence":["kalshi-refactor","kalshi-ack","kalshi-checks","kalshibot-catalog-1","kalshibot-catalog-2","kalshibot-catalog-3","kalshibot-catalog-4","kalshibot-catalog-5"],"projectIds":["kalshibot"],"tags":["kalshibot","Prediction-market execution systems","implementation","technical depth","architecture","result","Python","Reconciliation","Event-driven systems","Offline testing","Queue-aware replay","Qwen","OpenAI-compatible inference"]},{"id":"kalshibot-limits","kind":"FACT","text":"Prediction-market execution systems — evidence boundaries: ","evidence":["kalshi-refactor","kalshi-ack","kalshi-checks","kalshibot-catalog-1","kalshibot-catalog-2","kalshibot-catalog-3","kalshibot-catalog-4","kalshibot-catalog-5"],"projectIds":["kalshibot"],"tags":["kalshibot","Prediction-market execution systems","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"statistical-arbitrage-details","kind":"FACT","text":"Statistical arbitrage & module discovery: Dynamic peer graphs, efficient-price filtering, VECM residual experts, pair-quality admission, expected-net-alpha decomposition, portfolio constraints and a separation between search evaluation and promotion validation. Six constrained search tasks sit alongside graph and residual-model components, with separate promotion and full-system validation entrypoints.","evidence":["statistical-arbitrage-catalog-1","statistical-arbitrage-catalog-2","statistical-arbitrage-catalog-3","statistical-arbitrage-catalog-4"],"projectIds":["statistical-arbitrage"],"tags":["statistical-arbitrage","Statistical arbitrage & module discovery","implementation","technical depth","architecture","result","Python","Graphs","VECM","Statistical arbitrage","SkyDiscover","Deterministic evaluators"]},{"id":"statistical-arbitrage-limits","kind":"FACT","text":"Statistical arbitrage & module discovery — evidence boundaries: The design document mixes plans with implemented work; only source-backed components are claimed complete. Minute-bar search does not establish order-book execution fidelity. No gains from automated search are asserted.","evidence":["statistical-arbitrage-catalog-1","statistical-arbitrage-catalog-2","statistical-arbitrage-catalog-3","statistical-arbitrage-catalog-4"],"projectIds":["statistical-arbitrage"],"tags":["statistical-arbitrage","Statistical arbitrage & module discovery","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"options-variance-details","kind":"FACT","text":"Options & variance-risk-premium research: Black-Scholes pricing and delta solvers, chain/expiry selection, walk-forward protocols, overlap-aware variance evaluation, tail-risk filters and event-versus-diffusion variance with natural bid/ask fill comparisons. Reusable options research and execution components support walk-forward protocols and offline tests. The event study specifies prior-only jump estimates, shrinkage, and quote-quality gates.","evidence":["options-variance-catalog-1","options-variance-catalog-2","options-variance-catalog-3","options-variance-catalog-4"],"projectIds":["options-variance"],"tags":["options-variance","Options & variance-risk-premium research","implementation","technical depth","architecture","result","Python","IBKR API","Options pricing","Variance decomposition","Walk-forward research"]},{"id":"options-variance-limits","kind":"FACT","text":"Options & variance-risk-premium research — evidence boundaries: Not a claim of returns or live deployment. Simplified daily Black-Scholes marks differ from executable option quotes; natural-fill and mid-fill analyses must remain distinct. No broker/account state accessed.","evidence":["options-variance-catalog-1","options-variance-catalog-2","options-variance-catalog-3","options-variance-catalog-4"],"projectIds":["options-variance"],"tags":["options-variance","Options & variance-risk-premium research","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"lppls-research-cockpits-details","kind":"FACT","text":"LPPLS market-state research and interactive cockpits: Nested trailing-window LPPLS fits; purged out-of-fold meta-models and calibration; PIT entry/label clocks; frozen histories; incremental data refresh with rollback and regression guards; separating timing, stock selection, sizing and hedging; serving prebuilt research artifacts through FastAPI without recomputing the engine on each request. Inspectable dashboards expose market states, signals, simulated books, performance, and data freshness. Research records preserve rejected mechanisms and checks that stopped unsupported conclusions from advancing.","evidence":["lppls-research-cockpits-catalog-1","lppls-research-cockpits-catalog-2","lppls-research-cockpits-catalog-3","lppls-research-cockpits-catalog-4","lppls-research-cockpits-catalog-5","lppls-research-cockpits-catalog-6","lppls-research-cockpits-catalog-7","lppls-research-cockpits-catalog-8","lppls-research-cockpits-catalog-9"],"projectIds":["lppls-research-cockpits"],"tags":["lppls-research-cockpits","LPPLS market-state research and interactive cockpits","implementation","technical depth","architecture","result","Python","LPPLS","FastAPI","Purged cross-validation","Isotonic calibration","Parquet","Interactive dashboards"]},{"id":"lppls-research-cockpits-limits","kind":"FACT","text":"LPPLS market-state research and interactive cockpits — evidence boundaries: This pass inspected documentation and source only; it did not launch or validate current services, scheduled refresh, accounts or model results. Local documents use live for a current cockpit or paper workflow, which must not be presented as verified real-money execution. Historical return/Sortino/capacity figures are not promoted. Cost-basis histograms are vendor-imputed price/turnover proxies, not direct observations of investor intentions.","evidence":["lppls-research-cockpits-catalog-1","lppls-research-cockpits-catalog-2","lppls-research-cockpits-catalog-3","lppls-research-cockpits-catalog-4","lppls-research-cockpits-catalog-5","lppls-research-cockpits-catalog-6","lppls-research-cockpits-catalog-7","lppls-research-cockpits-catalog-8","lppls-research-cockpits-catalog-9"],"projectIds":["lppls-research-cockpits"],"tags":["lppls-research-cockpits","LPPLS market-state research and interactive cockpits","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"zero-human-hedge-details","kind":"FACT","text":"Zero Human Hedge — paper research operations: Research artifact manifests, session normalization, feature sweeps, constrained targets, backtest parity, DuckDB reporting, Prometheus/Grafana metrics and Paperclip integration. A paper-research operations stack with artifact manifests, Docker infrastructure, dashboards, and observability spanning the research loop.","evidence":["zero-human-hedge-catalog-1","zero-human-hedge-catalog-2","zero-human-hedge-catalog-3"],"projectIds":["zero-human-hedge"],"tags":["zero-human-hedge","Zero Human Hedge — paper research operations","implementation","technical depth","architecture","result","Python","DuckDB","NautilusTrader","Paperclip","Docker","Prometheus","Grafana"]},{"id":"zero-human-hedge-limits","kind":"FACT","text":"Zero Human Hedge — paper research operations — evidence boundaries: Paper fund is an experiment/operations metaphor, not evidence of managing a fund or client money. No autonomous profitability, deployed service status or external user claim.","evidence":["zero-human-hedge-catalog-1","zero-human-hedge-catalog-2","zero-human-hedge-catalog-3"],"projectIds":["zero-human-hedge"],"tags":["zero-human-hedge","Zero Human Hedge — paper research operations","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"mention-market-model-details","kind":"FACT","text":"World Cup announcer-mention modeling: Beta-smoothed word baselines, hierarchical crew/team effects, empirical-Bayes shrinkage, knockout-regime changes, bounded corpus-logit tilts and leave-one-out evaluation with corpus-leakage adjustments. The documented study found crew-settlement effects more informative than corpus tilts for words with settlement history. The result is a dated research finding, not an established trading edge.","evidence":["mention-market-model-catalog-1","mention-market-model-catalog-2","mention-market-model-catalog-3"],"projectIds":["mention-market-model"],"tags":["mention-market-model","World Cup announcer-mention modeling","implementation","technical depth","architecture","result","Python","NLP corpora","Empirical Bayes","Beta-binomial models","Calibration"]},{"id":"mention-market-model-limits","kind":"FACT","text":"World Cup announcer-mention modeling — evidence boundaries: Treat this as a dated July 2026 study. Captions and highlights have unequal coverage; first-mention timing is an approximation from price paths and excludes unsuitable contexts. No trading edge or out-of-sample deployment claim.","evidence":["mention-market-model-catalog-1","mention-market-model-catalog-2","mention-market-model-catalog-3"],"projectIds":["mention-market-model"],"tags":["mention-market-model","World Cup announcer-mention modeling","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"factor-replication-details","kind":"FACT","text":"A-share factor replication and evaluation: Announcement-date availability, monthly universes, market-specific tradability, factor preprocessing, cross-sectional neutralization, factor aggregation, portfolio formation and robust inference. The OpenAP harness covers a documented library of 212 predictor definitions, alongside ingestion, factor evaluation, and portfolio-formation workflows. Library scope is not a successful-factor count.","evidence":["factor-replication-catalog-1","factor-replication-catalog-2","factor-replication-catalog-3","factor-replication-catalog-4","factor-replication-catalog-5"],"projectIds":["factor-replication"],"tags":["factor-replication","A-share factor replication and evaluation","implementation","technical depth","architecture","result","Python","Tushare","Asset pricing","Factor models","PIT universes","Statistical inference","C++","HDF5","Market-data ingestion"]},{"id":"factor-replication-limits","kind":"FACT","text":"A-share factor replication and evaluation — evidence boundaries: 212 is the OpenAP library scope, not a verified successful-factor count. No profitable factor discovery, exact reproduction of every definition or Hikyuu authorship/release/adoption claim.","evidence":["factor-replication-catalog-1","factor-replication-catalog-2","factor-replication-catalog-3","factor-replication-catalog-4","factor-replication-catalog-5"],"projectIds":["factor-replication"],"tags":["factor-replication","A-share factor replication and evaluation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"mean-reversion-system-details","kind":"FACT","text":"A-share reversal with executable constraints: Lagged signals and next-open execution, overlapping holding periods, clustering-augmented reversals, continuous regime scaling, partial fills, futures beta hedges and factor-alpha diagnostics. The research system connects lagged signals to next-open execution, overlapping holdings, futures hedges, and factor-alpha diagnostics, with execution-simulator test coverage in source.","evidence":["mean-reversion-system-catalog-1","mean-reversion-system-catalog-2","mean-reversion-system-catalog-3"],"projectIds":["mean-reversion-system"],"tags":["mean-reversion-system","A-share reversal with executable constraints","implementation","technical depth","architecture","result","Python","Clustering","Mean reversion","Portfolio optimization","Execution simulation"]},{"id":"mean-reversion-system-limits","kind":"FACT","text":"A-share reversal with executable constraints — evidence boundaries: Research benchmark numbers in documentation are comparison targets, not achieved project returns. Exact historical execution and paper replication were not rerun; the design filename has an inconsistent year and should not determine public chronology.","evidence":["mean-reversion-system-catalog-1","mean-reversion-system-catalog-2","mean-reversion-system-catalog-3"],"projectIds":["mean-reversion-system"],"tags":["mean-reversion-system","A-share reversal with executable constraints","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"market-signal-ingestion-details","kind":"FACT","text":"Market-signal ingestion and replay: REST/WebSocket adapters, normalized source ingestion, reconnect/authentication handling, stream capture, replay and downstream alert interfaces. Normalized REST/WebSocket ingestion covers prediction markets, trading venues, and public disclosures, with reconnect handling and replayable streams for downstream research.","evidence":["market-signal-ingestion-catalog-1","market-signal-ingestion-catalog-2","market-signal-ingestion-catalog-3"],"projectIds":["market-signal-ingestion"],"tags":["market-signal-ingestion","Market-signal ingestion and replay","implementation","technical depth","architecture","result","Python","httpx","WebSockets","REST APIs","Replay","Dashboards"]},{"id":"market-signal-ingestion-limits","kind":"FACT","text":"Market-signal ingestion and replay — evidence boundaries: A downstream trading engine is explicitly outside this repository. No signal quality, commercial access entitlement or trading-performance claim. Do not expose session tokens or account integration details publicly.","evidence":["market-signal-ingestion-catalog-1","market-signal-ingestion-catalog-2","market-signal-ingestion-catalog-3"],"projectIds":["market-signal-ingestion"],"tags":["market-signal-ingestion","Market-signal ingestion and replay","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"technical-indicator-evaluation-details","kind":"FACT","text":"Technical-indicator evaluation workbench: Fold coverage planning, deterministic sample selection, indicator isolation, timeouts, failure replay, rank-information coefficients, cross-fold aggregation and checkpointed execution. An evaluator, API, and Streamlit workbench organize a documented library of 100 TradingView-derived indicators across repeatable folds and aggregated metrics.","evidence":["technical-indicator-evaluation-catalog-1","technical-indicator-evaluation-catalog-2"],"projectIds":["technical-indicator-evaluation"],"tags":["technical-indicator-evaluation","Technical-indicator evaluation workbench","implementation","technical depth","architecture","result","Python","pandas","NumPy","Streamlit","Parquet","Factor evaluation"]},{"id":"technical-indicator-evaluation-limits","kind":"FACT","text":"Technical-indicator evaluation workbench — evidence boundaries: ","evidence":["technical-indicator-evaluation-catalog-1","technical-indicator-evaluation-catalog-2"],"projectIds":["technical-indicator-evaluation"],"tags":["technical-indicator-evaluation","Technical-indicator evaluation workbench","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"market-mechanism-studies-details","kind":"FACT","text":"A-share event, flow and intervention studies: Availability clocks, event and era definitions, frozen panels, feasible-trade constraints, robust ETF-flow proxies, missing-feed confidence, and converting informal mechanisms into measurable tests. The studies preserve rejected hypotheses and test the limits of disclosure following. A seven-component tracker scores observable policy-support patterns, with its synthetic demonstration labeled explicitly.","evidence":["market-mechanism-studies-catalog-1","market-mechanism-studies-catalog-2","market-mechanism-studies-catalog-3","market-mechanism-studies-catalog-4","market-mechanism-studies-catalog-5","market-mechanism-studies-catalog-6"],"projectIds":["market-mechanism-studies"],"tags":["market-mechanism-studies","A-share event, flow and intervention studies","implementation","technical depth","architecture","result","Python","Event studies","Tushare","Time-series analysis","Hypothesis testing","Streamlit","Robust statistics","Interpretable scoring"]},{"id":"market-mechanism-studies-limits","kind":"FACT","text":"A-share event, flow and intervention studies — evidence boundaries: Report findings are dated and not independently reproduced. No financial figures promoted. Policy scores identify observable patterns, not actual buyer identity or intent. IPO listing returns do not establish achievable subscription returns.","evidence":["market-mechanism-studies-catalog-1","market-mechanism-studies-catalog-2","market-mechanism-studies-catalog-3","market-mechanism-studies-catalog-4","market-mechanism-studies-catalog-5","market-mechanism-studies-catalog-6"],"projectIds":["market-mechanism-studies"],"tags":["market-mechanism-studies","A-share event, flow and intervention studies","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"finance-reading-and-falsification-details","kind":"FACT","text":"Trading literature and falsification notebooks: Separating descriptive patterns from predictive claims, expressing rule availability, identifying tests that require unavailable data, outlining ablations and judging falsifiability. The notebooks separate descriptive patterns from predictive claims and identify which tests require unavailable evidence. The Al Brooks study produces a report and blueprint rather than an implemented strategy.","evidence":["finance-reading-and-falsification-catalog-1","finance-reading-and-falsification-catalog-2","finance-reading-and-falsification-catalog-3","finance-reading-and-falsification-catalog-4"],"projectIds":["finance-reading-and-falsification"],"tags":["finance-reading-and-falsification","Trading literature and falsification notebooks","implementation","technical depth","architecture","result","Literature synthesis","Experiment design","Falsification","Quantitative research"]},{"id":"finance-reading-and-falsification-limits","kind":"FACT","text":"Trading literature and falsification notebooks — evidence boundaries: Literature performance claims are external claims, not Donald’s results. Reading-material presence does not establish personal beliefs or endorsements. No full books or copied teaching material should be republished.","evidence":["finance-reading-and-falsification-catalog-1","finance-reading-and-falsification-catalog-2","finance-reading-and-falsification-catalog-3","finance-reading-and-falsification-catalog-4"],"projectIds":["finance-reading-and-falsification"],"tags":["finance-reading-and-falsification","Trading literature and falsification notebooks","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"csc-tactics-details","kind":"FACT","text":"CSC:tactics: Changing a drill changes scene objects, ball counts, and camera state together. Both renderers need consistent behavior across transitions. A complete teaching application connects tactical explanations to a 90-minute practice plan with 43 segments. September 2026 verification records 11 Node and 12 browser tests, including both renderers and every training segment.","evidence":["soccer-content","soccer-lifecycle","soccer-reflection","csc-tactics-catalog-1","csc-tactics-catalog-2","csc-tactics-catalog-3","csc-tactics-catalog-4"],"projectIds":["csc-tactics"],"tags":["csc-tactics","CSC:tactics","implementation","technical depth","architecture","result","JavaScript","Three.js","Canvas","Vite","Interactive learning"]},{"id":"csc-tactics-limits","kind":"FACT","text":"CSC:tactics — evidence boundaries: Screen learning is not validated improvement in match performance. No player/user counts or coaching biography established.","evidence":["soccer-content","soccer-lifecycle","soccer-reflection","csc-tactics-catalog-1","csc-tactics-catalog-2","csc-tactics-catalog-3","csc-tactics-catalog-4"],"projectIds":["csc-tactics"],"tags":["csc-tactics","CSC:tactics","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"youtube-knowledge-details","kind":"FACT","text":"YouTube Knowledge: Fast summary and deeper analysis are separated. A bounded semaphore permits three ingests, per-channel locks protect shared profiles, and temp-file replacement keeps writes atomic. NDJSON heartbeats keep long jobs observable. Graph validation restricts source video IDs and requires an actual cross-video connection; graph failure does not discard primary distillation. A Chinese-first application connects videos to summaries and cross-video ideas. Bounded ingestion, per-channel locking, atomic writes, and validated source IDs keep long-running knowledge workflows observable and recoverable.","evidence":["youtube-knowledge-catalog-1","youtube-knowledge-catalog-2","youtube-knowledge-catalog-3","youtube-knowledge-catalog-4","youtube-knowledge-catalog-5","youtube-knowledge-catalog-6"],"projectIds":["youtube-knowledge"],"tags":["youtube-knowledge","YouTube Knowledge","implementation","technical depth","architecture","result","Python","Claude CLI","NDJSON","Threads","Atomic file replacement","JSON","YouTube transcripts","JavaScript"]},{"id":"youtube-knowledge-limits","kind":"FACT","text":"YouTube Knowledge — evidence boundaries: Transcript content and creator conclusions are external material, not Donald-authored beliefs or writing. Uses Claude CLI for analysis; this is not local Qwen inference. Live uptime, extraction accuracy, user adoption, and all historical browser outcomes were not reverified. Only implementation and test source were inspected; stored transcripts, chat logs, and personal library contents were excluded.","evidence":["youtube-knowledge-catalog-1","youtube-knowledge-catalog-2","youtube-knowledge-catalog-3","youtube-knowledge-catalog-4","youtube-knowledge-catalog-5","youtube-knowledge-catalog-6"],"projectIds":["youtube-knowledge"],"tags":["youtube-knowledge","YouTube Knowledge","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"piano-midi-details","kind":"FACT","text":"Piano video to MIDI: Downloaded audio, model failures, GPU availability, overlapping notes, and variable timing need a coherent conversion flow. A conversion workflow with serial queueing, CPU fallback after CUDA failure, optional timing and velocity cleanup, overlap trimming, and a saved MIDI library.","evidence":["piano-pipeline","piano-midi-catalog-1","piano-midi-catalog-2","piano-midi-catalog-3","piano-midi-catalog-4"],"projectIds":["piano-midi"],"tags":["piano-midi","Piano video to MIDI","implementation","technical depth","architecture","result","Python","Flask","Transkun","MIDI","PyTorch","pretty-midi","yt-dlp"]},{"id":"piano-midi-limits","kind":"FACT","text":"Piano video to MIDI — evidence boundaries: Transkun is upstream, not a model Donald trained. No hand-splitting implementation found. The project does not establish that Donald plays piano or owns PianoVision. User library/audio files were not inspected. Transcription accuracy remains unmeasured.","evidence":["piano-pipeline","piano-midi-catalog-1","piano-midi-catalog-2","piano-midi-catalog-3","piano-midi-catalog-4"],"projectIds":["piano-midi"],"tags":["piano-midi","Piano video to MIDI","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"browser-repl-details","kind":"FACT","text":"Sandbox: browser code experiments: CodeMirror language extensions and completions; browser execution bridges and console serialization; lazy Pyodide/WebAssembly loading; worker lifecycle and execution timeouts; localStorage file management; formatting and export; draggable output layout. Two browser-only execution designs explore the same editor workflow, including lazy Pyodide loading and worker execution budgets. They are experiments rather than production-secure sandboxes.","evidence":["browser-repl-catalog-1","browser-repl-catalog-2","browser-repl-catalog-3","browser-repl-catalog-4"],"projectIds":["browser-repl"],"tags":["browser-repl","Sandbox: browser code experiments","implementation","technical depth","architecture","result","Vue","CodeMirror","Vite","Web Workers","Pyodide","WebAssembly"]},{"id":"browser-repl-limits","kind":"FACT","text":"Sandbox: browser code experiments — evidence boundaries: Origin is an evaluation/project folder; avoid implying paid employment or customer delivery. Do not call either variant a production-secure sandbox. The iframe/main-thread version cannot reliably interrupt blocking execution; worker version has explicit termination budgets. No authored commit history or usability/compatibility measurements verified.","evidence":["browser-repl-catalog-1","browser-repl-catalog-2","browser-repl-catalog-3","browser-repl-catalog-4"],"projectIds":["browser-repl"],"tags":["browser-repl","Sandbox: browser code experiments","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"daily-news-terminal-details","kind":"FACT","text":"Daily News Vision Terminal: A Node server normalizes heterogeneous public sources, labels source warnings and stale cache, scores textual relevance to markets, maintains keyword alerts, and pushes WebSocket updates to a globe and dense feed interface. A Docker-packaged server and frontend combine news, prediction-market context, trade activity, and flight overlays. Source warnings and stale-cache labels replace built-in fake-data fallbacks.","evidence":["daily-news-terminal-catalog-1","daily-news-terminal-catalog-2","daily-news-terminal-catalog-3","daily-news-terminal-catalog-4"],"projectIds":["daily-news-terminal"],"tags":["daily-news-terminal","Daily News Vision Terminal","implementation","technical depth","architecture","result","Node.js","JavaScript","WebSocket","Docker","GDELT","Polymarket public APIs","OpenSky ADS-B"]},{"id":"daily-news-terminal-limits","kind":"FACT","text":"Daily News Vision Terminal — evidence boundaries: Candidate-specific authorship is not corroborated by version history. Market matching is an entity/token heuristic, not demonstrated LLM reasoning or causal inference. No accuracy, reliability, live uptime, commercial adoption, or trading performance was established. Optional alerts and configured external integrations were not invoked.","evidence":["daily-news-terminal-catalog-1","daily-news-terminal-catalog-2","daily-news-terminal-catalog-3","daily-news-terminal-catalog-4"],"projectIds":["daily-news-terminal"],"tags":["daily-news-terminal","Daily News Vision Terminal","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"video-transcriber-adaptation-details","kind":"FACT","text":"Video transcription workflow adaptation: Async model-client compatibility layer collecting streamed text; model configuration normalization; subtitle-first versus audio fallback pipeline; GPU speech recognition; platform extraction and smoke scripts. The adaptation connects streamed model output and GPU speech recognition to an upstream subtitle-first workflow.","evidence":["video-transcriber-adaptation-catalog-1","video-transcriber-adaptation-catalog-2","video-transcriber-adaptation-catalog-3","video-transcriber-adaptation-catalog-4","video-transcriber-adaptation-catalog-5"],"projectIds":["video-transcriber-adaptation"],"tags":["video-transcriber-adaptation","Video transcription workflow adaptation","implementation","technical depth","architecture","result","Python","FastAPI","Faster-Whisper","CUDA","yt-dlp","Claude Agent SDK"]},{"id":"video-transcriber-adaptation-limits","kind":"FACT","text":"Video transcription workflow adaptation — evidence boundaries: Upstream app authorship remains with its original contributors. Model-list strings are implementation configuration, not independent confirmation of provider product availability. No live platform extraction, model call or GPU benchmark was run. Transcript/media temp files and credentials were not inspected. Throughput and transcription accuracy have not been benchmarked.","evidence":["video-transcriber-adaptation-catalog-1","video-transcriber-adaptation-catalog-2","video-transcriber-adaptation-catalog-3","video-transcriber-adaptation-catalog-4","video-transcriber-adaptation-catalog-5"],"projectIds":["video-transcriber-adaptation"],"tags":["video-transcriber-adaptation","Video transcription workflow adaptation","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"personal-website-details","kind":"FACT","text":"A searchable engineering notebook: A common knowledge model for pages and machine-readable feeds; schema-constrained claim selection; server-side inference; signed answer snapshots; persistent sharing; reduced-motion spatial interaction. An implemented website with field navigation, searchable projects, detailed evidence, recruiter views, JSON feeds and a local-Qwen answer service.","evidence":["personal-website-catalog-1","personal-website-catalog-2","personal-website-catalog-3"],"projectIds":["personal-website"],"tags":["personal-website","A searchable engineering notebook","implementation","technical depth","architecture","result","React","TypeScript","Cloudflare","Local Qwen","Semantic HTML"]},{"id":"personal-website-limits","kind":"FACT","text":"A searchable engineering notebook — evidence boundaries: AI-assisted implementation. The guide selects reviewed claims rather than reproducing Donald’s voice; missing biographical information remains unknown.","evidence":["personal-website-catalog-1","personal-website-catalog-2","personal-website-catalog-3"],"projectIds":["personal-website"],"tags":["personal-website","A searchable engineering notebook","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"technology-society-writing-details","kind":"FACT","text":"Cars, identity & technological change: Evidence-based synthesis across a film and historical readings; distinguishing the promises of technological progress from its social effects; an academic writing artifact that broadens the engineering picture. An authored essay examining how transport technology shapes social life and identity, broadening the collection beyond engineering implementation.","evidence":["technology-society-writing-catalog-1","technology-society-writing-catalog-2"],"projectIds":["technology-society-writing"],"tags":["technology-society-writing","Cars, identity & technological change","implementation","technical depth","architecture","result","Academic writing","Technology history","LaTeX"]},{"id":"technology-society-writing-limits","kind":"FACT","text":"Cars, identity & technological change — evidence boundaries: Do not infer favorite films, political views, personal motivations or identity traits from the essay. Do not republish long excerpts from quoted books or the film. Fordlandia exists as a PDF but its authorship/content was not inspected; list as a related local artifact only.","evidence":["technology-society-writing-catalog-1","technology-society-writing-catalog-2"],"projectIds":["technology-society-writing"],"tags":["technology-society-writing","Cars, identity & technological change","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"illinix-391-details","kind":"FACT","text":"Illinix 391: ELF binary loading, preemptive multitasking, UNIX pipes, and shell I/O redirection connect process execution, scheduling, storage, and device I/O into one operating system. A complete kernel supporting process creation, executable loading, filesystem access, multitasking, and an interactive shell.","evidence":["illinix-391-catalog-1"],"projectIds":["illinix-391"],"tags":["illinix-391","Illinix 391","implementation","technical depth","architecture","result","C","RISC-V Assembly","Virtual memory","Operating systems","VIRTIO","KTFS"]},{"id":"illinix-391-limits","kind":"FACT","text":"Illinix 391 — evidence boundaries: Project architecture and scope are from Donald’s Spring 2025 project summary.","evidence":["illinix-391-catalog-1"],"projectIds":["illinix-391"],"tags":["illinix-391","Illinix 391","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"reaction-wheel-pendulum-details","kind":"FACT","text":"Reaction Wheel Pendulum: The controller combines a nonlinear mechanical model, state feedback, observer-based state estimation, and real-time actuation through Wincon. Achieved stable equilibrium control of the unstable nonlinear system.","evidence":["reaction-wheel-pendulum-catalog-1"],"projectIds":["reaction-wheel-pendulum"],"tags":["reaction-wheel-pendulum","Reaction Wheel Pendulum","implementation","technical depth","architecture","result","Control systems","Lagrangian dynamics","State feedback","Luenberger observer","Wincon"]},{"id":"reaction-wheel-pendulum-limits","kind":"FACT","text":"Reaction Wheel Pendulum — evidence boundaries: Control outcome and implementation details are from Donald’s Fall 2024 project summary.","evidence":["reaction-wheel-pendulum-catalog-1"],"projectIds":["reaction-wheel-pendulum"],"tags":["reaction-wheel-pendulum","Reaction Wheel Pendulum","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"twentyfourpoints-details","kind":"FACT","text":"twentyfourpoints.com: Mathematical validation algorithms and auto-balancing support competitive play. The delivery includes unit tests, migration scripts, and cross-device deployment. Launched in 72 hours. Donald’s Summer 2025 project record reports 760+ unique puzzles solved and 100% uptime during the reported launch period.","evidence":["twentyfourpoints-catalog-1"],"projectIds":["twentyfourpoints"],"tags":["twentyfourpoints","twentyfourpoints.com","implementation","technical depth","architecture","result","React","TypeScript","Node.js","Socket.io","Supabase","ELO"]},{"id":"twentyfourpoints-limits","kind":"FACT","text":"twentyfourpoints.com — evidence boundaries: Launch time, puzzle count, and uptime are candidate-reported Summer 2025 outcomes. The uptime observation interval was not specified; it is not a current service guarantee. Puzzle counts are not user counts.","evidence":["twentyfourpoints-catalog-1"],"projectIds":["twentyfourpoints"],"tags":["twentyfourpoints","twentyfourpoints.com","limitations","boundaries","deployment","accuracy","authorship","compiler","runtime"]},{"id":"field-robotics","kind":"FACT","text":"Robotics & Autonomous Driving: Autonomous driving lab — A ROS2/Gazebo course project connecting lane segmentation, curvature-aware speed control, and LiDAR/GPS localization. Reaction Wheel Pendulum — Real-time stabilization of an inherently unstable pendulum using a reaction wheel. Multi-robot navigation — A potential-field controller evaluated over every goal assignment for five robots in a fixed obstacle-filled circular workspace.","evidence":["autonomous-driving-lab-catalog-1","reaction-wheel-pendulum-catalog-1","swarm-code"],"projectIds":["autonomous-driving-lab","reaction-wheel-pendulum","swarm-navigation"],"tags":["Robotics & Autonomous Driving","Robotics & autonomy","robotics"]},{"id":"field-embedded-vision","kind":"FACT","text":"Embedded Systems & Computer Vision: Driver monitoring & diagnostic lab — On-device driver-state monitoring developed during Donald’s research-engineering internship at Shanghai Ben’an Intelligent: eight detector classes, vehicle-aware alerts, and replay-driven validation. DMS on CV181x — A C inference pipeline for constrained hardware: sparse face detection, tracked regions, lightweight landmarks, and per-stage timing. VGGT scene reconstruction — A local adaptation of VGGT that filters sky regions from reconstructed point clouds and adds reconstruction inspection tools.","evidence":["benan-dms","benan-dms","vggt-reconstruction-catalog-1"],"projectIds":["driver-monitoring","dms-cv181x","vggt-reconstruction"],"tags":["Embedded Systems & Computer Vision","Embedded systems & vision","embedded-vision"]},{"id":"field-machine-learning","kind":"FACT","text":"Machine Learning & Model Research: Time-series forecasting & retrieval — Time-series foundation models meet nearest-neighbor retrieval, causal evaluation, and ablations that expose prediction-to-decision failures. Momentum Transformer & execution learning — An extensive adaptation of the upstream Momentum Transformer research, including a PyTorch implementation and portfolio/execution reinforcement-learning layers. mini_torch — autograd from NumPy — A NumPy-only neural-network training library with dynamic computation graphs, reverse-mode differentiation and a small MNIST classifier.","evidence":["chronos-architecture","momentum-transformer-adaptation-catalog-1","mini-torch-catalog-1"],"projectIds":["chronos-knn","momentum-transformer-adaptation","mini-torch"],"tags":["Machine Learning & Model Research","Machine learning","machine-learning"]},{"id":"field-local-llms","kind":"FACT","text":"Local LLM Inference & Agents: Local Qwen inference — Self-hosted Qwen workflows with controlled request budgets, reproducible decoding, structured decisions, and explicit inference-failure handling. ZhengMindYAGNI — agent orchestration — An AI-work coordinator built around Git, markdown work nodes, a small state machine, and a CLI that owns mutations. VTA — local Qwen adaptation — A local adaptation of an upstream financial time-series LLM pipeline to Qwen3.5, spanning inference compatibility and staged GRPO/LoRA/SFT training code.","evidence":["local-qwen-inference-catalog-1","zhengmind-yagni-catalog-1","vta-qwen-adaptation-catalog-1"],"projectIds":["local-qwen-inference","zhengmind-yagni","vta-qwen-adaptation"],"tags":["Local LLM Inference & Agents","Local LLMs & agents","local-llms"]},{"id":"field-software-systems","kind":"FACT","text":"Software Systems & Automation: Illinix 391 — A Unix-like kernel in C and RISC-V assembly, from virtual memory to an interactive shell. ISAFE / WITNESS — SOP verification — An assembly-inspection prototype that tracks procedure steps, flags skips and ordering errors, and attaches video evidence for review. A Go-style runtime in C++ — Cooperative scheduling, epoll-backed I/O, timers, and synchronization in C++17. Waiting tasks yield instead of blocking a worker.","evidence":["illinix-391-catalog-1","isafe-sop-catalog-1","cpp-goroutine-catalog-1"],"projectIds":["illinix-391","isafe-sop","cpp-goroutine"],"tags":["Software Systems & Automation","Systems & automation","software-systems"]},{"id":"field-quant-finance","kind":"FACT","text":"Quantitative Finance & Markets: Prediction-market execution systems — Order reconciliation, execution controls, queue-aware research, and an independently packaged refactor of a prediction-market system. Statistical arbitrage & module discovery — Graph-conditioned residuals, pair selection, and a constrained search harness for testing alternative research modules. Options & variance-risk-premium research — Options research and execution tooling spanning implied-versus-realized volatility, short-strangle protocols, and an earnings-event variance decomposition study.","evidence":["kalshi-refactor","statistical-arbitrage-catalog-1","options-variance-catalog-1"],"projectIds":["kalshibot","statistical-arbitrage","options-variance"],"tags":["Quantitative Finance & Markets","Quantitative finance","quant-finance"]},{"id":"field-creative-tools","kind":"FACT","text":"Interactive Tools, Learning & Writing: twentyfourpoints.com — A real-time multiplayer math arena, built and launched in 72 hours. CSC:tactics — A Chinese soccer tactics teaching application with 52 situational challenges, 11 roles, three viewpoints and a 90-minute training plan containing 43 segments. YouTube Knowledge — A Chinese-first reading and exploration application that turns video transcripts into staged summaries, source-linked ideas, and creator-level knowledge graphs.","evidence":["twentyfourpoints-catalog-1","soccer-content","youtube-knowledge-catalog-1"],"projectIds":["twentyfourpoints","csc-tactics","youtube-knowledge"],"tags":["Interactive Tools, Learning & Writing","Tools, learning & writing","creative-tools"]},{"id":"built-overview","kind":"FACT","text":"Donald’s work spans autonomous driving and robotics, embedded driver monitoring, machine learning, local LLM inference and agents, industrial SOP verification, software systems, quantitative finance, and interactive learning tools. The index organizes the work into seven fields, with related repositories grouped and ranked within each field.","evidence":["autonomous-driving-lab-catalog-1","driver-monitoring-catalog-1","isafe-sop-catalog-1","local-qwen-inference-catalog-1","chronos-architecture","kalshi-refactor","soccer-content"],"projectIds":["autonomous-driving-lab","driver-monitoring","isafe-sop","local-qwen-inference","chronos-knn","kalshibot","csc-tactics"],"tags":["built","projects","overview","work"]},{"id":"forecast-failure","kind":"FACT","text":"In a forecasting experiment, optimizing trend clarity made nearly every forecast positive. The simulator exited only on a negative value, so positions accumulated. It is a documented failure of the prediction-to-decision chain, not a live trading loss claim.","evidence":["chronos-failure"],"projectIds":["chronos-knn"],"tags":["failed","failure","wrong","loss","negative","experiment"]},{"id":"attention-failure","kind":"FACT","text":"In the X-Trend reproduction, both attention keys and values initially encoded conditions alone. The correction adds observed returns to the values. Current source contains that change.","evidence":["xtrend-values","xtrend-source"],"projectIds":["xtrend"],"tags":["attention","deep","tensor","research","ml"]},{"id":"execution-depth","kind":"FACT","text":"The Kalshibot refactor explicitly handles ambiguous order creation: an accepted order may exist even when the network response is lost. Its request path avoids automatically retrying that non-idempotent operation.","evidence":["kalshi-ack","kalshi-refactor"],"projectIds":["kalshibot"],"tags":["depth","engineering","system","production","risk","execution","strongest"]},{"id":"swarm-negative","kind":"FACT","text":"The swarm-navigation report documents orbiting, local minima, and collisions during parameter exploration. Excessive goal attraction could overpower repulsion. These are simulator findings.","evidence":["swarm-failure"],"projectIds":["swarm-navigation"],"tags":["robotics","swarm","failure"]},{"id":"soccer-negative","kind":"FACT","text":"In CSC:tactics, switching from a two-ball drill to one ball left a stale extra-ball mesh. A camera reset rendered before scene synchronization. Synchronizing the meshes before the reset fixed the transition.","evidence":["soccer-lifecycle"],"projectIds":["csc-tactics"],"tags":["soccer","failure"]},{"id":"swarm-outcome","kind":"FACT","text":"The saved swarm-navigation results show 120 of 120 goal assignments completing in the fixed simulation suite. This does not establish general convergence or physical-robot performance.","evidence":["swarm-results"],"projectIds":["swarm-navigation"],"tags":["result","robot","hardware","simulation","120"]},{"id":"recurring-pattern","kind":"SUPPORTED INTERPRETATION","text":"A recurring connection in the work is the gap between a promising model and the behavior of the system around it: forecast exits, order acknowledgements, and scene transitions all make that gap concrete.","evidence":["chronos-failure","kalshi-ack","soccer-lifecycle"],"projectIds":["chronos-knn","kalshibot","csc-tactics"],"tags":["think","approach","why","markets","problems","good","care","interests","pattern"]},{"id":"current-focus","kind":"FACT","text":"Donald’s current work includes the DMS research-engineering internship at Shanghai Ben’an Intelligent (2025–present, remote). Fall 2026 study includes AI Systems & Engineering and High Frequency Trading Technology, alongside Digital Systems Laboratory, Applied Parallel Programming, and Introduction to Optimization.","evidence":["benan-dms","academic-foundation"],"projectIds":["driver-monitoring","dms-cv181x"],"tags":["now","current","recent","obsessed","learning"]},{"id":"detour","kind":"FACT","text":"A less expected project is a piano-video-to-MIDI converter: it wraps upstream Transkun transcription with timing cleanup, velocity smoothing, queueing, and a saved library. The code does not establish that Donald plays piano.","evidence":["piano-pipeline"],"projectIds":["piano-midi"],"tags":["weird","unexpected","fun","music","piano","hobby","detour"]},{"id":"soccer-note","kind":"FACT","text":"The CSC:tactics training plan ends with players identifying an effective combination they noticed. That small reflection prompt sits alongside 43 practical training segments.","evidence":["soccer-reflection","soccer-content"],"projectIds":["csc-tactics"],"tags":["human","soccer","football","teaching","team"]},{"id":"interview","kind":"SUPPORTED INTERPRETATION","text":"Useful interview prompts include: How would you reconcile an order with a lost acknowledgement? Why did a forecasting objective disable exits? How do you verify that attention values contain the information the paper requires?","evidence":["kalshi-ack","chronos-failure","xtrend-source"],"projectIds":["kalshibot","chronos-knn","xtrend"],"tags":["interview","evaluate","hire","ask","questions"]},{"id":"research-boundary","kind":"FACT","text":"Donald lists two 2021 IEEE conference publications: Vehicle Trajectory Prediction with Goal Estimation (CEI) and Visual Analytics for the International Trade (ICVISP). His research spans autonomous-driving trajectory prediction and mathematical modeling/data visualization for trade. Both papers have linked DOI records and bibliographic details.","evidence":["trajectory-publication","trade-publication"],"projectIds":[],"tags":["paper","publication","published","research","academic"]},{"id":"unknown-personal","kind":"UNKNOWN","text":"The notebook does not contain verified personal motivations, team preferences, favorite books or music, or a first-person writing corpus. Donald is the right person to answer those directly.","evidence":[],"projectIds":[],"tags":["believe","values","favorite","personality","motivation","opinion","team preferences"]},{"id":"unknown-career","kind":"UNKNOWN","text":"For availability, graduation timing, work authorization, or a current résumé, contact Donald directly. Academic performance information is private.","evidence":[],"projectIds":[],"tags":["resume","graduation","gpa","internship","employment","availability","visa","salary","citizenship"]},{"id":"unknown-impact","kind":"UNKNOWN","text":"Donald’s math arena has candidate-reported launch outcomes: 760+ unique puzzles solved and 100% uptime during the reported Summer 2025 launch period. Puzzle counts do not establish user counts. Trading profitability, company impact, and external ML customer counts require separate evidence.","evidence":["twentyfourpoints-catalog-1"],"projectIds":["twentyfourpoints"],"tags":["profit","return","pnl","users","customers","production ml","deployed","impact"]},{"id":"unknown","kind":"UNKNOWN","text":"There is not enough reviewed evidence in this notebook to answer that reliably. You can ask Donald directly, or explore the project notes and their stated limits.","evidence":[],"projectIds":[],"tags":[]}],"evidence":[{"id":"identity","title":"Candidate-provided identity","projectId":"","source":"Website brief supplied by Donald","date":"2026-09-17","excerpt":"Donald Shen · donald7@illinois.edu · Computer Engineering at UIUC","context":"Name, academic field, and contact were supplied directly. Professional experience and project records are provided separately.","kind":"candidate","url":"/evidence/identity"},{"id":"kalshi-refactor","title":"An independent execution system","projectId":"kalshibot","source":"kalshibot/docs/refactor.md","date":"2026-09-04","excerpt":"An independent source snapshot separates execution, research, and mutable runtime state. Structural equivalence does not establish production equivalence.","context":"Paraphrase of the refactor record. Donald-attributed Git commits establish work on the independent refactor. Agent assistance is documented; original upstream and operating-system modules are not claimed as unaided work.","kind":"report","url":"/evidence/kalshi-refactor"},{"id":"kalshi-ack","title":"The lost-acknowledgement case","projectId":"kalshibot","source":"kalshibot/tests/trading/test_membrane_fixes.py:726","date":"2026-09-17","excerpt":"A non-idempotent POST must NOT be retried on an ambiguous 5xx","context":"Source comment, corroborated by client request handling and the lost-response test. An order may exist even if its response did not arrive. These are implementation safeguards, not evidence of profitability.","kind":"source","url":"/evidence/kalshi-ack"},{"id":"kalshi-checks","title":"What the verification actually establishes","projectId":"kalshibot","source":"kalshibot/docs/verification.md","date":"2026-09-04","excerpt":"The dated verification record documents 1,526 passing tests with network access blocked, plus fresh-wheel installation checks.","context":"Paraphrase of historical documentation. This website audit did not rerun that suite. Live-account takeover, exact production replay, and economic equivalence remain separate gates.","kind":"report","url":"/evidence/kalshi-checks"},{"id":"chronos-architecture","title":"Retrieval with a deliberately simple baseline","projectId":"chronos-knn","source":"chronos_knn/docs/architecture/README.md","date":"2026-09-17","excerpt":"Twenty-day windows are represented by normalized raw features and retrieved through Qdrant, with versioned preprocessing and evaluation boundaries.","context":"Paraphrase of architecture and source. Git records replacement of Chronos embeddings with normalized raw features in February 2026. Chronos, Kronos, and Granite TTM are upstream models.","kind":"source","url":"/evidence/chronos-architecture"},{"id":"chronos-failure","title":"The exit that never happened","projectId":"chronos-knn","source":"chronos_knn/report/ceiling_implementation/ablation_executive_summary.md:29","date":"2026-03-06","excerpt":"The exit mechanism (tc < 0) never fires","context":"The experimental trend-clarity objective produced positive, smooth forecasts for almost every stock. A simulated exit rule required a negative value, so positions accumulated. This is a documented simulation failure, not a claim of live losses.","kind":"report","url":"/evidence/chronos-failure"},{"id":"chronos-regime","title":"A hypothesis the study did not establish","projectId":"chronos-knn","source":"chronos_knn/report/ttm_channel_mix_regime/executive_summary.md","date":"2026-03","excerpt":"The study did not establish an improvement from weekly regime-selected retraining over the rolling baseline.","context":"Conservative paraphrase of the recorded ablation. Overlapping intervals alone are not a full significance test. Reported investment returns are intentionally not used as proof.","kind":"report","url":"/evidence/chronos-regime"},{"id":"xtrend-values","title":"Conditions are not outcomes","projectId":"xtrend","source":"xtrend_revised/BUG_FIX_SUMMARY_KEY_VALUE_ENCODING.md:7","date":"2025-11","excerpt":"Keys should match on MARKET CONDITIONS. Values should contain OUTCOMES.","context":"Project debugging note. The initial implementation encoded both using features alone. The corrected value projection concatenates context features and observed returns.","kind":"report","url":"/evidence/xtrend-values"},{"id":"xtrend-source","title":"The correction in the tensor","projectId":"xtrend","source":"xtrend_revised/xtrend/models/qkv_projections.py:80","date":"2026-09-17","excerpt":"torch.cat([context_states, context_returns_expanded], dim=-1)","context":"Current source excerpt. Git history attributes implementation and fixes to Donald, with agent-assisted commits also present. This is a reproduction/adaptation of the X-Trend approach, not original authorship of its paper.","kind":"source","url":"/evidence/xtrend-source"},{"id":"soccer-content","title":"One pitch, eleven perspectives","projectId":"csc-tactics","source":"soccer/README.md","date":"2026-09","excerpt":"52 tactical challenges, 11 selectable roles, three camera viewpoints, and eight drills containing 43 training segments.","context":"Counts checked against current project documentation and drill data. A Chinese teaching application built with JavaScript, Three.js, and a Canvas fallback. No player-outcome or adoption claim is implied.","kind":"source","url":"/evidence/soccer-content"},{"id":"soccer-lifecycle","title":"A camera change exposed a stale scene","projectId":"csc-tactics","source":"soccer/docs/4231_测试与限制.md","date":"2026-09-12","excerpt":"Scene meshes must be synchronized before resetCamera triggers an immediate render.","context":"Paraphrase of the documented failure: a two-ball drill transitioned to one ball while an extra-ball mesh remained. The repair synchronized scene state before rendering. Historical browser checks are recorded; not newly rerun in this audit.","kind":"report","url":"/evidence/soccer-lifecycle"},{"id":"soccer-reflection","title":"A small piece of the training plan","projectId":"csc-tactics","source":"soccer/src/data/drills.json:248","date":"2026-09-17","excerpt":"The final drill asks players to name an effective combination they noticed.","context":"English paraphrase of Chinese app content. This is an artifact from the teaching tool, not a verified personal quotation or evidence of Donald’s role on a team.","kind":"source","url":"/evidence/soccer-reflection"},{"id":"swarm-code","title":"Potential fields, with somewhere to go","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/student_code/my_nav_fn.py","date":"2025-12-01","excerpt":"Goal attraction, obstacle repulsion and circulation, robot repulsion, and conditional boundary repulsion form the navigation controller.","context":"Source summary. Course report names Donald Shen, and Donald-attributed commits modify the controller. The surrounding simulation envelope is supplied coursework infrastructure.","kind":"source","url":"/evidence/swarm-code"},{"id":"swarm-results","title":"120 assignments, checked in the saved data","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/results/summary.csv","date":"2026-09-17","excerpt":"120 rows. 120 success=True. Mean completion: 8,434.06 simulation steps.","context":"Independently recomputed from saved results during the website audit. This is the fixed simulation suite, not a general convergence guarantee or a physical-robot result. Inconsistent extrema in the prose report are not used.","kind":"data","url":"/evidence/swarm-results"},{"id":"swarm-failure","title":"When attraction gets in the way","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/report.md","date":"2025-12","excerpt":"The report records orbiting, local minima, and collisions when goal gains overpower repulsion.","context":"Paraphrase of course failure analysis. Suggested future work includes deadlock detection, velocity-aware avoidance, and adaptive gains. These suggestions are not presented as implemented.","kind":"report","url":"/evidence/swarm-failure"},{"id":"piano-pipeline","title":"A video becomes editable notes","projectId":"piano-midi","source":"pianovision_midi_converter/pipeline/postprocessor.py","date":"2026-09-17","excerpt":"MIDI processing estimates tempo, moves timing 80% toward a sixteenth-note grid, smooths velocity, and trims same-pitch overlaps.","context":"Current implementation summary. Donald-attributed February 2026 commit introduces the converter. Transkun performs upstream audio transcription; the project adds ingestion, processing, a queue and a saved library. Transcription accuracy is not established.","kind":"source","url":"/evidence/piano-pipeline"},{"id":"control-stop","title":"No trustworthy path, no motion command","projectId":"f1tenth","source":"uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws/src/f1tenth_control/f1tenth_control/pure_pursuit_control.py","date":"2025-11","excerpt":"Camera-only mode stops when no valid path is available.","context":"Source behavior and Donald-attributed safety-fix commits. The pipeline uses camera lane extraction, a path selector, and pure pursuit in an existing F1Tenth ROS2 workspace. Physical track performance was not verified.","kind":"source","url":"/evidence/control-stop"},{"id":"benan-dms","title":"Shanghai Ben’an Intelligent · Research Engineer, Driver State Monitoring (DMS)","projectId":"","source":"Candidate-provided internship experience and deployment update","date":"2025 – Present","excerpt":"Deployed and validated real-time on-device driver-state monitoring, connecting safety-relevant alerts, low-power edge inference, and replay-driven validation. Deployed and validated eight safety-relevant detector classes: eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly. Shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics alongside model development. Designed the alert state machine with hierarchical face-loss and lens-occlusion fallbacks. Speed, ignition, and gear gating suppress false alarms in non-driving states; persisted per-camera calibration profiles support fleet-grade traceability. Fit the model pipeline to the low-power CV181x edge SoC by tightening inference cadence and per-stage compute budgets. Validated end-to-end on real cabin video and built the lab CLI for shadow-mode hard-case capture, human review, and replay-based metric reporting.","context":"Role, period, and contribution supplied directly by Donald.","kind":"candidate","url":"/evidence/benan-dms"},{"id":"sjtu-research","title":"Shanghai Jiao Tong University · Autonomous Driving Research","projectId":"","source":"Candidate-provided experience","date":"2021","excerpt":"Implemented YOLOv3 object detection, OpenCV and SLAM, and lane-following algorithms on a ROS platform for autonomous navigation.","context":"Role, period, and contribution supplied directly by Donald.","kind":"candidate","url":"/evidence/sjtu-research"},{"id":"mckinsey-research","title":"McKinsey & Company · Part-Time Research Analyst","projectId":"","source":"Candidate-provided experience","date":"2021","excerpt":"Produced McKinsey-standard research reports on industrial automation and inventory management systems.","context":"Role, period, and contribution supplied directly by Donald.","kind":"candidate","url":"/evidence/mckinsey-research"},{"id":"berkeley-summer","title":"University of California, Berkeley · Summer Session, EECS","projectId":"","source":"Candidate-provided education experience","date":"Summer 2022","excerpt":"Summer Session, EECS · Summer 2022. Summer study in electrical engineering and computer sciences.","context":"Summer-session attendance supplied directly by Donald. This is an education experience, not a degree claim.","kind":"candidate","url":"/evidence/berkeley-summer"},{"id":"trajectory-publication","title":"Vehicle Trajectory Prediction with Goal Estimation","projectId":"","source":"IEEE · 10.1109/CEI52496.2021.9574469","url":"/evidence/trajectory-publication","date":"2021","excerpt":"Vehicle Trajectory Prediction with Goal Estimation. Chenxi Jin; Jiadong Shen. 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI), pp. 474–479.","context":"Bibliography verified against IEEE-deposited Crossref metadata; contribution supplied by Donald.","kind":"report","sourceUrl":"https://doi.org/10.1109/CEI52496.2021.9574469"},{"id":"trade-publication","title":"Visual Analytics for the International Trade","projectId":"","source":"IEEE · 10.1109/ICVISP54630.2021.00059","url":"/evidence/trade-publication","date":"2021","excerpt":"Visual Analytics for the International Trade. Haowen Jiang; Jiadong Shen; Qiujie Chou; Ziling Dong; Shenghui Cheng. 2021 5th International Conference on Vision, Image and Signal Processing (ICVISP), pp. 296–301.","context":"Bibliography verified against IEEE-deposited Crossref metadata; contribution supplied by Donald.","kind":"report","sourceUrl":"https://doi.org/10.1109/ICVISP54630.2021.00059"},{"id":"academic-foundation","title":"Engineering and scientific coursework","projectId":"","source":"Candidate-supplied academic record and current enrollment update","date":"2026-09-17","excerpt":"Course names and completion status support the academic foundation shown on the background page.","context":"Completed courses, current study, and transfer or examination credit are distinguished. Current enrollment reflects Donald’s latest update; completed coursework remains based on the academic record. The source document remains private.","kind":"candidate","url":"/evidence/academic-foundation"},{"id":"autonomous-driving-lab-catalog-1","title":"run_lane_detection.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/src/mp1/scripts/run_lane_detection.py","date":"2026-09-17","excerpt":"Donald commit 7d45ffc modifies lane-segmentation steps; current node combines segmentation, BEV geometry and fitted lane output.","url":"/evidence/autonomous-driving-lab-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"autonomous-driving-lab-catalog-2","title":"controller.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/src/mp2/src/controller.py","date":"2026-09-17","excerpt":"Donald commit c31b8d0 adds curvature-aware longitudinal control; later candidate commits refine the course controller.","url":"/evidence/autonomous-driving-lab-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"autonomous-driving-lab-catalog-3","title":"particle_filter.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/src/mp3/src/particle_filter.py","date":"2026-09-17","excerpt":"Donald commit 2f1b315 implements likelihood, normalization, resampling and GPS-related logic.","url":"/evidence/autonomous-driving-lab-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"autonomous-driving-lab-catalog-4","title":"tune_particle_filter.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/src/mp3/tools/tune_particle_filter.py","date":"2026-09-17","excerpt":"Donald commit f39ba27 adds standalone generated-sequence parameter sweeps and explicit evaluation metrics.","url":"/evidence/autonomous-driving-lab-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"autonomous-driving-lab-catalog-5","title":"lidar_processing.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/src/mp3/src/lidar_processing.py","date":"2026-09-17","excerpt":"Extended LiDAR measurement processing in the attributed localization commit.","url":"/evidence/autonomous-driving-lab-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"autonomous-driving-lab-catalog-6","title":"dubin_controller.py","projectId":"autonomous-driving-lab","source":"uiuc/ece484/mp0/dubin_controller.py","date":"2026-09-17","excerpt":"Preliminary Dubins-mode decision logic with scaffold boundaries.","url":"/evidence/autonomous-driving-lab-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"swarm-navigation-catalog-1","title":"report.md","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/report.md","date":"2026-09-17","excerpt":"Donald-named report, method, parameter sensitivity, failures and future improvements.","url":"/evidence/swarm-navigation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"swarm-navigation-catalog-2","title":"my_nav_fn.py","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/student_code/my_nav_fn.py","date":"2026-09-17","excerpt":"Implemented controller; Donald commits 5f01b8a and 21eab11, December 2025.","url":"/evidence/swarm-navigation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"swarm-navigation-catalog-3","title":"summary.csv","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/results/summary.csv","date":"2026-09-17","excerpt":"120 recorded successes and independently recomputed mean.","url":"/evidence/swarm-navigation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"swarm-navigation-catalog-4","title":"gen_failure.py","projectId":"swarm-navigation","source":"uiuc/ECE_470_FA25_Project-Code/gen_failure.py","date":"2026-09-17","excerpt":"Failure-case visualization tooling.","url":"/evidence/swarm-navigation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"f1tenth-catalog-1","title":"README.md","projectId":"f1tenth","source":"uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws/README.md","date":"2026-09-17","excerpt":"Identifies the upstream base ROS2 workspace.","url":"/evidence/f1tenth-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"f1tenth-catalog-2","title":"lane_detector_node.py","projectId":"f1tenth","source":"uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws/src/f1tenth_control/f1tenth_control/lane_detector_node.py","date":"2026-09-17","excerpt":"Camera extraction and homography-axis conversion; commits 83501d4 and 79dcab5 attribute additions to Donald.","url":"/evidence/f1tenth-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"f1tenth-catalog-3","title":"pure_pursuit_control.py","projectId":"f1tenth","source":"uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws/src/f1tenth_control/f1tenth_control/pure_pursuit_control.py","date":"2026-09-17","excerpt":"Missing dynamic path and lookahead paths publish stop behavior; Donald safety-fix commits f751b48/f597280.","url":"/evidence/f1tenth-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"f1tenth-catalog-4","title":"camera-lane-following-testing-context.md","projectId":"f1tenth","source":"uiuc/ece484_f1_tenth_cr7_ws/f1_tenth_cr7_ws/dev/active/camera-lane-following-testing/camera-lane-following-testing-context.md","date":"2026-09-17","excerpt":"Pipeline design and initial validation boundaries.","url":"/evidence/f1tenth-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"ur3-manipulation-catalog-1","title":"lab2_exec.py","projectId":"ur3-manipulation","source":"uiuc/ece470/lab/lab2_exec.py","date":"2026-09-17","excerpt":"Recursive three-block planning, gripper sensing, arm movement and stateful block transfer.","url":"/evidence/ur3-manipulation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"ur3-manipulation-catalog-2","title":"blob_search.py","projectId":"ur3-manipulation","source":"uiuc/ece470/lab/lab5pkg_py/scripts/blob_search.py","date":"2026-09-17","excerpt":"Implemented HSV filtering, morphological cleanup, blob criteria and coordinate conversion; calibration remains placeholder.","url":"/evidence/ur3-manipulation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"ur3-manipulation-catalog-3","title":"lab5_func.py","projectId":"ur3-manipulation","source":"uiuc/ece470/lab/lab5pkg_py/scripts/lab5_func.py","date":"2026-09-17","excerpt":"Unimplemented forward/inverse-kinematics sections identify the current boundary.","url":"/evidence/ur3-manipulation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"ur3-manipulation-catalog-4","title":"lab5_exec.py","projectId":"ur3-manipulation","source":"uiuc/ece470/lab/lab5pkg_py/scripts/lab5_exec.py","date":"2026-09-17","excerpt":"Pick/place orchestration with empty calibrated goal lists.","url":"/evidence/ur3-manipulation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"driver-monitoring-catalog-1","title":"README.md","projectId":"driver-monitoring","source":"dms/README.md","date":"2026-09-17","excerpt":"Architecture, eight detection states, calibration workflow, shared lab and historical test count.","url":"/evidence/driver-monitoring-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"driver-monitoring-catalog-2","title":"types.py","projectId":"driver-monitoring","source":"dms/src/dms/types.py","date":"2026-09-17","excerpt":"FrameContext and VehicleContext explicitly carry source timestamps and motion-alert gating.","url":"/evidence/driver-monitoring-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"driver-monitoring-catalog-3","title":"registry.py","projectId":"driver-monitoring","source":"dms/src/dms/detectors/registry.py","date":"2026-09-17","excerpt":"Secondary hand-inference scheduling and shared detector orchestration.","url":"/evidence/driver-monitoring-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"driver-monitoring-catalog-4","title":"summary.md","projectId":"driver-monitoring","source":"dms/fix/260401-2050-detector-contracts/summary.md","date":"2026-09-17","excerpt":"Documented smoking-latch, stale-obstruction and early-warning fixes.","url":"/evidence/driver-monitoring-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"driver-monitoring-catalog-5","title":"replay.py","projectId":"driver-monitoring","source":"dms/src/dms/lab/replay.py","date":"2026-09-17","excerpt":"Diagnostic replay implementation; sample data was not inspected.","url":"/evidence/driver-monitoring-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"dms-cv181x-catalog-1","title":"README.md","projectId":"dms-cv181x","source":"dms_cv181x_fastpath/dms_work/README.md","date":"2026-09-17","excerpt":"Historical architecture and bundle-status record, including hardware checks that were still open at that snapshot. Donald’s later internship update describes on-device deployment and real cabin-video validation.","url":"/evidence/dms-cv181x-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"dms-cv181x-catalog-2","title":"cvi_dms_service.c","projectId":"dms-cv181x","source":"dms_cv181x_fastpath/dms_work/src/cvi_dms_service.c","date":"2026-09-17","excerpt":"Sparse detector scheduling, tracked ROI, landmark backends and latency fields.","url":"/evidence/dms-cv181x-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"dms-cv181x-catalog-3","title":"README.md","projectId":"dms-cv181x","source":"dms_cv181x_fastpath/dms_work/artifacts/model_bundle/package/README.md","date":"2026-09-17","excerpt":"Vendor SCRFD provenance, public PFLD ONNX provenance and BF16 conversion notes.","url":"/evidence/dms-cv181x-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"dms-cv181x-catalog-4","title":"SCRUTINY_AND_MIGRATION.md","projectId":"dms-cv181x","source":"dms_cv181x_fastpath/dms_work/docs/SCRUTINY_AND_MIGRATION.md","date":"2026-09-17","excerpt":"Critical-path and API problems motivating the refactor; model availability statements are older than the main README.","url":"/evidence/dms-cv181x-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vggt-reconstruction-catalog-1","title":"reconstruct_3d_no_sky.py","projectId":"vggt-reconstruction","source":"vggt/reconstruct_3d_no_sky.py","date":"2026-09-17","excerpt":"Adds sky segmentation and filtering around upstream VGGT inference and geometry utilities.","url":"/evidence/vggt-reconstruction-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vggt-reconstruction-catalog-2","title":"check_results.py","projectId":"vggt-reconstruction","source":"vggt/check_results.py","date":"2026-09-17","excerpt":"Local output-inspection utility.","url":"/evidence/vggt-reconstruction-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vggt-reconstruction-catalog-3","title":"viser_stable_no_sky.py","projectId":"vggt-reconstruction","source":"vggt/viser_stable_no_sky.py","date":"2026-09-17","excerpt":"Local visualization of filtered reconstructions.","url":"/evidence/vggt-reconstruction-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"sam3-experiments-catalog-1","title":"sam3_image_predictor_example.ipynb","projectId":"sam3-experiments","source":"sam3/examples/sam3_image_predictor_example.ipynb","date":"2026-09-17","excerpt":"Supplied predictor notebook with a locally changed text prompt and executed artifacts.","url":"/evidence/sam3-experiments-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"sam3-experiments-catalog-2","title":"pyproject.toml","projectId":"sam3-experiments","source":"sam3/pyproject.toml","date":"2026-09-17","excerpt":"Local Python compatibility adjustment against the upstream package.","url":"/evidence/sam3-experiments-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"chronos-knn-catalog-1","title":"README.md","projectId":"chronos-knn","source":"chronos_knn/docs/architecture/README.md","date":"2026-09-17","excerpt":"Normalized raw-feature retrieval, Qdrant and schema/version boundaries.","url":"/evidence/chronos-knn-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"chronos-knn-catalog-2","title":"ablation_executive_summary.md","projectId":"chronos-knn","source":"chronos_knn/report/ceiling_implementation/ablation_executive_summary.md","date":"2026-09-17","excerpt":"March 2026 negative ablation: the exit mechanism stopped firing.","url":"/evidence/chronos-knn-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"chronos-knn-catalog-3","title":"persistence_head.py","projectId":"chronos-knn","source":"chronos_knn/src/chronos_knn/forecast/persistence_head.py","date":"2026-09-17","excerpt":"Classifier on TTM decoder hidden states; March 10 Donald-attributed training/inference commits.","url":"/evidence/chronos-knn-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"chronos-knn-catalog-4","title":"model.md","projectId":"chronos-knn","source":"A_Share/kronos_rag/model.md","date":"2026-09-17","excerpt":"Canonical snapback factor is mean reversion against stale Kronos envelopes, not fresh forecast following.","url":"/evidence/chronos-knn-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"momentum-transformer-adaptation-catalog-1","title":"README.md","projectId":"momentum-transformer-adaptation","source":"trading-momentum-transformer/README.md","date":"2026-09-17","excerpt":"Explicit upstream paper authorship and local-data entrypoints.","url":"/evidence/momentum-transformer-adaptation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"momentum-transformer-adaptation-catalog-2","title":"execution_agent.py","projectId":"momentum-transformer-adaptation","source":"trading-momentum-transformer/mom_trans_pytorch/rl/agents/execution_agent.py","date":"2026-09-17","excerpt":"Implemented base, threshold-signal and RL execution agents with trade projection.","url":"/evidence/momentum-transformer-adaptation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"momentum-transformer-adaptation-catalog-3","title":"execution_trading_env.py","projectId":"momentum-transformer-adaptation","source":"trading-momentum-transformer/mom_trans_pytorch/rl/envs/execution_trading_env.py","date":"2026-09-17","excerpt":"Adapter for reset, step, observation conversion and fills.","url":"/evidence/momentum-transformer-adaptation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"momentum-transformer-adaptation-catalog-4","title":"Contribution history","projectId":"momentum-transformer-adaptation","source":"trading-momentum-transformer/.git","date":"2026-09-17","excerpt":"Donald-attributed commits 0f35813, 7ec8ce7, 9aaa3dc and 38cca49 establish adaptation and corrected PPO/cost behavior.","url":"/evidence/momentum-transformer-adaptation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mini-torch-catalog-1","title":"README.md","projectId":"mini-torch","source":"uiuc/ece449/CS446_ECE449_SP2026_MP1/README.md","date":"2026-09-17","excerpt":"Defines the supplied NumPy-only autograd assignment and starter boundaries.","url":"/evidence/mini-torch-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mini-torch-catalog-2","title":"ops.py","projectId":"mini-torch","source":"uiuc/ece449/CS446_ECE449_SP2026_MP1/mini_torch/ops.py","date":"2026-09-17","excerpt":"Implemented Function.apply, unbroadcasting, derivatives and cross-entropy.","url":"/evidence/mini-torch-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mini-torch-catalog-3","title":"tensor.py","projectId":"mini-torch","source":"uiuc/ece449/CS446_ECE449_SP2026_MP1/mini_torch/tensor.py","date":"2026-09-17","excerpt":"Reverse topological traversal, accumulation and graph cleanup.","url":"/evidence/mini-torch-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mini-torch-catalog-4","title":"layers.py","projectId":"mini-torch","source":"uiuc/ece449/CS446_ECE449_SP2026_MP1/mini_torch/nn/layers.py","date":"2026-09-17","excerpt":"Trainable Linear layer implementation.","url":"/evidence/mini-torch-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mini-torch-catalog-5","title":"mnist_classification.py","projectId":"mini-torch","source":"uiuc/ece449/CS446_ECE449_SP2026_MP1/mnist_classification.py","date":"2026-09-17","excerpt":"Implemented MLP, training-data normalization and SGD training loop.","url":"/evidence/mini-torch-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-1","title":"README.md","projectId":"rl-portfolio-framework","source":"rl_trader_framework/README.md","date":"2026-09-17","excerpt":"Signal contract, holdings-aware environment, constrained PPO and regime-source distinction.","url":"/evidence/rl-portfolio-framework-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-2","title":"ppo_pid.py","projectId":"rl-portfolio-framework","source":"rl_trader_framework/rl_trader/agents/ppo_pid.py","date":"2026-09-17","excerpt":"Concrete constrained RL agent module.","url":"/evidence/rl-portfolio-framework-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-3","title":"trading_env.py","projectId":"rl-portfolio-framework","source":"rl_trader_framework/rl_trader/envs/trading_env.py","date":"2026-09-17","excerpt":"Concrete holdings and cost environment.","url":"/evidence/rl-portfolio-framework-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-4","title":"README.md","projectId":"rl-portfolio-framework","source":"lstm_ppo_regime_starter/README.md","date":"2026-09-17","excerpt":"RecurrentPPO, expanding-window GMM and long-only-with-cash baseline.","url":"/evidence/rl-portfolio-framework-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-5","title":"README.md","projectId":"rl-portfolio-framework","source":"tet-jumpmodels-tushare/README.md","date":"2026-09-17","excerpt":"Technical TET signals, jumpmodels and logistic overlay.","url":"/evidence/rl-portfolio-framework-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-6","title":"README.md","projectId":"rl-portfolio-framework","source":"tet_fundamentals_tushare_a_share/README.md","date":"2026-09-17","excerpt":"Availability-lag alignment and three-arm ablation design.","url":"/evidence/rl-portfolio-framework-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-7","title":"a_share_fundamentals.py","projectId":"rl-portfolio-framework","source":"tet_fundamentals_tushare_a_share/src/fundamentals/a_share_fundamentals.py","date":"2026-09-17","excerpt":"Quarterly/TTM and robust cross-sectional transforms implemented.","url":"/evidence/rl-portfolio-framework-catalog-7","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rl-portfolio-framework-catalog-8","title":"2402.05272v1.pdf","projectId":"rl-portfolio-framework","source":"jump_model_regime/2402.05272v1.pdf","date":"2026-09-17","excerpt":"Reference artifact only; paper contents not independently audited in this pass.","url":"/evidence/rl-portfolio-framework-catalog-8","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"xtrend-catalog-1","title":"BUG_FIX_SUMMARY_KEY_VALUE_ENCODING.md","projectId":"xtrend","source":"xtrend_revised/BUG_FIX_SUMMARY_KEY_VALUE_ENCODING.md","date":"2026-09-17","excerpt":"Values initially omitted observed outcomes; debugging note explains the correction.","url":"/evidence/xtrend-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"xtrend-catalog-2","title":"qkv_projections.py","projectId":"xtrend","source":"xtrend_revised/xtrend/models/qkv_projections.py","date":"2026-09-17","excerpt":"Outcome concatenation is present in current value projection.","url":"/evidence/xtrend-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"xtrend-catalog-3","title":"CPD_NORMALIZATION_FIX_SUMMARY.md","projectId":"xtrend","source":"xtrend_revised/CPD_NORMALIZATION_FIX_SUMMARY.md","date":"2026-09-17","excerpt":"Historical normalization failure and threshold limitations.","url":"/evidence/xtrend-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"xtrend-catalog-4","title":"README.md","projectId":"xtrend","source":"XTREND/README.md","date":"2026-09-17","excerpt":"Earlier A-share few-shot implementation, dashboard and market-friction scope.","url":"/evidence/xtrend-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"xtrend-catalog-5","title":"Contribution history","projectId":"xtrend","source":"xtrend_revised/.git","date":"2026-09-17","excerpt":"4ff852e records causal normalization; 5f5ed8a records expanding-window training.","url":"/evidence/xtrend-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-1","title":"mp1_starter.ipynb","projectId":"generative-models","source":"uiuc/ece598/mp1/mp1_starter.ipynb","date":"2026-09-17","excerpt":"Implemented convolutional VAE reparameterization, loss and multiscale MMD evaluation.","url":"/evidence/generative-models-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-2","title":"mp2_starter.ipynb","projectId":"generative-models","source":"uiuc/ece598/MP2/MP2/mp2_starter.ipynb","date":"2026-09-17","excerpt":"Implemented DDPM epsilon loss, classifier-free conditional/unconditional loss and samplers.","url":"/evidence/generative-models-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-3","title":"dps.ipynb","projectId":"generative-models","source":"uiuc/ece598/MP3/dps.ipynb","date":"2026-09-17","excerpt":"Implemented Tweedie estimate and simplified observation-guidance update inside supplied diffusion code.","url":"/evidence/generative-models-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-4","title":"dps copy.ipynb","projectId":"generative-models","source":"uiuc/ece598/MP3/dps copy.ipynb","date":"2026-09-17","excerpt":"Extended experiment/report-generation notebook.","url":"/evidence/generative-models-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-5","title":"report_template.md","projectId":"generative-models","source":"uiuc/ece598/MP3/part_b_results/report_template.md","date":"2026-09-17","excerpt":"Filled Donald-named November 17, 2025 report; six guidance strengths and four noise levels.","url":"/evidence/generative-models-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"generative-models-catalog-6","title":"README.md","projectId":"generative-models","source":"uiuc/ece598/MP3/guided-diffusion/README.md","date":"2026-09-17","excerpt":"Upstream model/code boundary.","url":"/evidence/generative-models-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"adaptive-portfolio-learning-catalog-1","title":"README.md","projectId":"adaptive-portfolio-learning","source":"A_Share/doubleadapt_stockmixer/README.md","date":"2026-09-17","excerpt":"Monthly DoubleAdapt protocol, causal labels, simulator and promotion layers.","url":"/evidence/adaptive-portfolio-learning-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"adaptive-portfolio-learning-catalog-2","title":"README.md","projectId":"adaptive-portfolio-learning","source":"A_Share/FSD/README.md","date":"2026-09-17","excerpt":"Typed PIT contracts, executable simulator, feasible teacher and reliability-gated imitation-learning scaffold.","url":"/evidence/adaptive-portfolio-learning-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"adaptive-portfolio-learning-catalog-3","title":"simulator.py","projectId":"adaptive-portfolio-learning","source":"A_Share/FSD/src/lr_fsd/simulator.py","date":"2026-09-17","excerpt":"Concrete simulation module, alongside teacher, DAgger, training and reliability modules.","url":"/evidence/adaptive-portfolio-learning-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"adaptive-portfolio-learning-catalog-4","title":"WALKFORWARD_VERDICT.md","projectId":"adaptive-portfolio-learning","source":"A_Share/deepm/docs/research/WALKFORWARD_VERDICT.md","date":"2026-09-17","excerpt":"Dated adverse result for a tested long-only selection variant.","url":"/evidence/adaptive-portfolio-learning-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"adaptive-portfolio-learning-catalog-5","title":"Contribution history","projectId":"adaptive-portfolio-learning","source":"A_Share/deepm/.git","date":"2026-09-17","excerpt":"52877f5 attributes the A-share cost-aware and neutralized experiments to donald7.","url":"/evidence/adaptive-portfolio-learning-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"project-genji-catalog-1","title":"README.md","projectId":"project-genji","source":"donald_trading_model/README.md","date":"2026-09-17","excerpt":"Original system roadmap and data-ingestion scope; early status is stale.","url":"/evidence/project-genji-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"project-genji-catalog-2","title":"train.py","projectId":"project-genji","source":"donald_trading_model/src/model2/train.py","date":"2026-09-17","excerpt":"Concrete CV splitter, trainers, fold metrics and OOF aggregation.","url":"/evidence/project-genji-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"project-genji-catalog-3","title":"features.py","projectId":"project-genji","source":"donald_trading_model/src/model2/features.py","date":"2026-09-17","excerpt":"Industry normalization, winsorization and multi-family features.","url":"/evidence/project-genji-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"project-genji-catalog-4","title":"Contribution history","projectId":"project-genji","source":"donald_trading_model/.git","date":"2026-09-17","excerpt":"b91cb79 and 2b15753 establish CV logging and multi-horizon training; 1a1db3d records determinism coverage.","url":"/evidence/project-genji-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"symbolic-alpha-mining-catalog-1","title":"README.md","projectId":"symbolic-alpha-mining","source":"A_Share/factorResearch/AlphaPROBE/README.md","date":"2026-09-17","excerpt":"Names the original AlphaPROBE authors and approach.","url":"/evidence/symbolic-alpha-mining-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"symbolic-alpha-mining-catalog-2","title":"pit.py","projectId":"symbolic-alpha-mining","source":"A_Share/factorResearch/AlphaPROBE/src/alphagen/data/pit.py","date":"2026-09-17","excerpt":"Local added PIT layer identified in Git status.","url":"/evidence/symbolic-alpha-mining-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"symbolic-alpha-mining-catalog-3","title":"feature_registry.py","projectId":"symbolic-alpha-mining","source":"A_Share/factorResearch/AlphaPROBE/src/alphagen/data/feature_registry.py","date":"2026-09-17","excerpt":"Local feature-registry extension identified alongside allocation additions.","url":"/evidence/symbolic-alpha-mining-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-1","title":"readme.md","projectId":"visual-alpha-research","source":"A_Share/factorResearch/CNN/readme.md","date":"2026-09-17","excerpt":"Chart rendering, CNN baseline, feature controls, fusion and evaluation research layout.","url":"/evidence/visual-alpha-research-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-2","title":"README.md","projectId":"visual-alpha-research","source":"A_Share/factorResearch/vit-sdf-lasso/README.md","date":"2026-09-17","excerpt":"Explicit A-share adaptation and active workflow boundaries.","url":"/evidence/visual-alpha-research-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-3","title":"master_variants.py","projectId":"visual-alpha-research","source":"A_Share/factorResearch/MASTER/adapters/master_variants.py","date":"2026-09-17","excerpt":"Local adaptation source exists on the upstream MASTER checkout.","url":"/evidence/visual-alpha-research-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-4","title":"dataset.py","projectId":"visual-alpha-research","source":"A_Share/factorResearch/MASTER/adapters/dataset.py","date":"2026-09-17","excerpt":"Local dataset adapter accompanies feature and market-input modules.","url":"/evidence/visual-alpha-research-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-5","title":"SPEC.md","projectId":"visual-alpha-research","source":"A_Share/attention_factors/SPEC.md","date":"2026-09-17","excerpt":"Upstream attribution and separate reference/deployable-construction designs.","url":"/evidence/visual-alpha-research-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-6","title":"train.py","projectId":"visual-alpha-research","source":"A_Share/attention_factors/src/afsa/train.py","date":"2026-09-17","excerpt":"Training module exists with model, cost and backtest code.","url":"/evidence/visual-alpha-research-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"visual-alpha-research-catalog-7","title":"variant_b2.py","projectId":"visual-alpha-research","source":"A_Share/attention_factors/src/afsa/variant_b2.py","date":"2026-09-17","excerpt":"Alternative portfolio construction source.","url":"/evidence/visual-alpha-research-catalog-7","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rotation-learning-catalog-1","title":"goal.md","projectId":"rotation-learning","source":"A_Share/Rotation/goal.md","date":"2026-09-17","excerpt":"Defines an event-level ranking question rather than independent stock forecasts.","url":"/evidence/rotation-learning-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rotation-learning-catalog-2","title":"README.md","projectId":"rotation-learning","source":"A_Share/Rotation/HMM/README.md","date":"2026-09-17","excerpt":"Filtered Student-t HMM, choice calibration, sizing and PIT data clock modules.","url":"/evidence/rotation-learning-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rotation-learning-catalog-3","title":"README.md","projectId":"rotation-learning","source":"A_Share/Rotation/gnn_solution/README.md","date":"2026-09-17","excerpt":"Dynamic latent rotation graph and exclusion of non-PIT industry classifications.","url":"/evidence/rotation-learning-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rotation-learning-catalog-4","title":"README.md","projectId":"rotation-learning","source":"A_Share/Rotation/graph_solution/README.md","date":"2026-09-17","excerpt":"Non-ML signed graph updates only when future windows mature.","url":"/evidence/rotation-learning-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"rotation-learning-catalog-5","title":"goal.md","projectId":"rotation-learning","source":"A_Share/Rotation_hmm_v14_bg/goal.md","date":"2026-09-17","excerpt":"Same core question supports grouping as a related variant.","url":"/evidence/rotation-learning-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cnn-ablation-harness-catalog-1","title":"README.md","projectId":"cnn-ablation-harness","source":"uiuc/cs441/README.md","date":"2026-09-17","excerpt":"Explicit PyTorch tutorial reproduction and step-to-flag map.","url":"/evidence/cnn-ablation-harness-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cnn-ablation-harness-catalog-2","title":"config.py","projectId":"cnn-ablation-harness","source":"uiuc/cs441/cnn_quiz/config.py","date":"2026-09-17","excerpt":"Experiment configuration.","url":"/evidence/cnn-ablation-harness-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cnn-ablation-harness-catalog-3","title":"data.py","projectId":"cnn-ablation-harness","source":"uiuc/cs441/cnn_quiz/data.py","date":"2026-09-17","excerpt":"Data transform variants.","url":"/evidence/cnn-ablation-harness-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cnn-ablation-harness-catalog-4","title":"train.py","projectId":"cnn-ablation-harness","source":"uiuc/cs441/cnn_quiz/train.py","date":"2026-09-17","excerpt":"Training/evaluation structure.","url":"/evidence/cnn-ablation-harness-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cnn-ablation-harness-catalog-5","title":"test_steps.py","projectId":"cnn-ablation-harness","source":"uiuc/cs441/tests/test_steps.py","date":"2026-09-17","excerpt":"Step wiring tests.","url":"/evidence/cnn-ablation-harness-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"upstream-ai-reference-library-catalog-1","title":"README.md","projectId":"upstream-ai-reference-library","source":"CAMEF/README.md","date":"2026-09-17","excerpt":"Explicit original authors and KDD 2025 paper attribution.","url":"/evidence/upstream-ai-reference-library-catalog-1","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"upstream-ai-reference-library-catalog-2","title":"Contribution history","projectId":"upstream-ai-reference-library","source":"CAMEF/.git","date":"2026-09-17","excerpt":"Clean lakebodhi/CAMEF upstream checkout with no Donald-attributed changes.","url":"/evidence/upstream-ai-reference-library-catalog-2","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"upstream-ai-reference-library-catalog-3","title":"README.md","projectId":"upstream-ai-reference-library","source":"kalshi/README.md","date":"2026-09-17","excerpt":"Upstream bot documentation.","url":"/evidence/upstream-ai-reference-library-catalog-3","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"upstream-ai-reference-library-catalog-4","title":"Contribution history","projectId":"upstream-ai-reference-library","source":"kalshi/.git","date":"2026-09-17","excerpt":"Origin ryanfrigo/kalshi-ai-trading-bot and clean worktree; no Donald commits.","url":"/evidence/upstream-ai-reference-library-catalog-4","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"upstream-ai-reference-library-catalog-5","title":"README.md","projectId":"upstream-ai-reference-library","source":"prediction_market/kalshi_wc_mention_analysis/README.md","date":"2026-09-17","excerpt":"Explicit statement that the local mention study was kept outside the cloned bot.","url":"/evidence/upstream-ai-reference-library-catalog-5","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"local-qwen-inference-catalog-1","title":"qwen.py","projectId":"local-qwen-inference","source":"kalshibot-membrane/autoresearch/lab/qwen.py","date":"2026-09-17","excerpt":"InferenceBudget reservations, fixed decoding, model_manifest, schema output, target-scaled completion budgets, and explicit failure handling.","url":"/evidence/local-qwen-inference-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"local-qwen-inference-catalog-2","title":"README.md","projectId":"local-qwen-inference","source":"market-ending/README.md","date":"2026-09-17","excerpt":"Private local Qwen pipeline and measured end-to-end timing descriptions.","url":"/evidence/local-qwen-inference-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"local-qwen-inference-catalog-3","title":"market-ending git history: d7e9855","projectId":"local-qwen-inference","source":"market-ending git history: d7e9855","date":"2026-09-17","excerpt":"Donald-attributed configuration of the existing local/home vLLM service.","url":"/evidence/local-qwen-inference-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"local-qwen-inference-catalog-4","title":"qwen_vl.py","projectId":"local-qwen-inference","source":"aisop/packages/aisop/adjudication/clients/qwen_vl.py","date":"2026-09-17","excerpt":"Complementary local multimodal inference path with 4-bit quantization and lazy loading.","url":"/evidence/local-qwen-inference-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-1","title":"README.md","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI/README.md","date":"2026-09-17","excerpt":"Git/files source of truth, dependency-aware work states, lessons, and agent commands.","url":"/evidence/zhengmind-yagni-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-2","title":"lock.go","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI/internal/core/lock.go","date":"2026-09-17","excerpt":"Exclusive file creation is the concurrent claim boundary.","url":"/evidence/zhengmind-yagni-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-3","title":"state.go","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI/internal/core/state.go","date":"2026-09-17","excerpt":"Small explicit state-transition model.","url":"/evidence/zhengmind-yagni-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-4","title":"spawn.go","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI/internal/web/spawn.go","date":"2026-09-17","excerpt":"Local agent processes and CLI mutation seam.","url":"/evidence/zhengmind-yagni-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-5","title":"architecture-review.md","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI/docs/architecture-review.md","date":"2026-09-17","excerpt":"Documented reconsideration of governance complexity and proposal to simplify around existing primitives.","url":"/evidence/zhengmind-yagni-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-yagni-catalog-6","title":"ZhengMindYAGNI git history: 103fb21, 226394f, 6055a89","projectId":"zhengmind-yagni","source":"ZhengMindYAGNI git history: 103fb21, 226394f, 6055a89","date":"2026-09-17","excerpt":"Donald-attributed simplification, reusable lessons, and streamed loop observability.","url":"/evidence/zhengmind-yagni-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vta-qwen-adaptation-catalog-1","title":"grpo_utils.py","projectId":"vta-qwen-adaptation","source":"A_Share/VTA/grpo/grpo_utils.py","date":"2026-09-17","excerpt":"Qwen3.5 detection, typed messages, dataset normalization, model loading and inference compatibility.","url":"/evidence/vta-qwen-adaptation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vta-qwen-adaptation-catalog-2","title":"requirements-qwen35.txt","projectId":"vta-qwen-adaptation","source":"A_Share/VTA/grpo/requirements-qwen35.txt","date":"2026-09-17","excerpt":"Pinned local CUDA/PyTorch/Transformers profile and explicit Unsloth inference route.","url":"/evidence/vta-qwen-adaptation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vta-qwen-adaptation-catalog-3","title":"stage1_train_grpo.py","projectId":"vta-qwen-adaptation","source":"A_Share/VTA/grpo/stage1_train_grpo.py","date":"2026-09-17","excerpt":"GRPO setup around shared model utilities.","url":"/evidence/vta-qwen-adaptation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vta-qwen-adaptation-catalog-4","title":"README.md","projectId":"vta-qwen-adaptation","source":"A_Share/VTA/grpo/README.md","date":"2026-09-17","excerpt":"Training-stage architecture, with outcome claims requiring separate validation.","url":"/evidence/vta-qwen-adaptation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"vta-qwen-adaptation-catalog-5","title":"VTA working-tree diff","projectId":"vta-qwen-adaptation","source":"A_Share/VTA working-tree diff","date":"2026-09-17","excerpt":"Scoped current modifications against the clean upstream import; no candidate-authored commit attribution found for this path.","url":"/evidence/vta-qwen-adaptation-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"qwen-market-ending-catalog-1","title":"README.md","projectId":"qwen-market-ending","source":"market-ending/README.md","date":"2026-09-17","excerpt":"Pipeline, distinctions between facts and prices, target moneyline exclusion, small dated benchmark, and limits.","url":"/evidence/qwen-market-ending-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"qwen-market-ending-catalog-2","title":"snapshot.py","projectId":"qwen-market-ending","source":"market-ending/snapshot.py","date":"2026-09-17","excerpt":"Gathering, formatting, single/multi-horizon inference, and evaluation implementation.","url":"/evidence/qwen-market-ending-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"qwen-market-ending-catalog-3","title":"qwen_judge.py","projectId":"qwen-market-ending","source":"market-ending/experiments/qwen_judge.py","date":"2026-09-17","excerpt":"Earlier facts-only model/rule comparison.","url":"/evidence/qwen-market-ending-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"qwen-market-ending-catalog-4","title":"market-ending git history: d7e9855, 8e214a7","projectId":"qwen-market-ending","source":"market-ending git history: d7e9855, 8e214a7","date":"2026-09-17","excerpt":"Mixed contributor history; Donald serving configuration and jd prompt revision.","url":"/evidence/qwen-market-ending-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"semantic-research-agents-catalog-1","title":"README.md","projectId":"semantic-research-agents","source":"A_Share/tradingagents/README.md","date":"2026-09-17","excerpt":"Explicit A-share semantic-alpha fork with six research boxes and structured long-only actions.","url":"/evidence/semantic-research-agents-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"semantic-research-agents-catalog-2","title":"workflow.py","projectId":"semantic-research-agents","source":"A_Share/tradingagents/tradingagents/structural_alpha/workflow.py","date":"2026-09-17","excerpt":"Local specialized research workflow source and Donald-attributed commits.","url":"/evidence/semantic-research-agents-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"semantic-research-agents-catalog-3","title":"Contribution history","projectId":"semantic-research-agents","source":"A_Share/atlas/.git","date":"2026-09-17","excerpt":"Donald-attributed prompt-evolution and research changes alongside upstream authors.","url":"/evidence/semantic-research-agents-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"semantic-research-agents-catalog-4","title":"README.md","projectId":"semantic-research-agents","source":"A_Share/ashare_semantic_long/ashare_semantic_long_project/README.md","date":"2026-09-17","excerpt":"Semantic confirmation, governance/accounting vetoes and constrained portfolio pipeline.","url":"/evidence/semantic-research-agents-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"agent-context-blueprint-catalog-1","title":"README.md","projectId":"agent-context-blueprint","source":"claude-context-blueprint-template/README.md","date":"2026-09-17","excerpt":"Planner/skeletoner/implementer/integrator workflow and artifact-based handoffs.","url":"/evidence/agent-context-blueprint-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"agent-context-blueprint-catalog-2","title":"post_patch_pipeline.py","projectId":"agent-context-blueprint","source":"claude-context-blueprint-template/.claude/hooks/post_patch_pipeline.py","date":"2026-09-17","excerpt":"Ticket validation, patch handling, Codex review gate, and integration commands.","url":"/evidence/agent-context-blueprint-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"agent-context-blueprint-catalog-3","title":"subagent_workflow.md","projectId":"agent-context-blueprint","source":"claude-context-blueprint-template/subagent_workflow.md","date":"2026-09-17","excerpt":"Workflow conventions and handoff structure.","url":"/evidence/agent-context-blueprint-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"agent-context-blueprint-catalog-4","title":"claude-context-blueprint-template git history: 5af92db, ce105eb, dc8b6c9","projectId":"agent-context-blueprint","source":"claude-context-blueprint-template git history: 5af92db, ce105eb, dc8b6c9","date":"2026-09-17","excerpt":"Donald-attributed workflow, branch-policy, and initial implementation commits.","url":"/evidence/agent-context-blueprint-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"claude-code-autopilot-catalog-1","title":"README.md","projectId":"claude-code-autopilot","source":"claude-code-autopilot/README.md","date":"2026-09-17","excerpt":"Explicit upstream attribution, activation behavior, reusable hooks, installation design.","url":"/evidence/claude-code-autopilot-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"claude-code-autopilot-catalog-2","title":"install.sh","projectId":"claude-code-autopilot","source":"claude-code-autopilot/install.sh","date":"2026-09-17","excerpt":"Installer implementation and backup/verification flow.","url":"/evidence/claude-code-autopilot-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"claude-code-autopilot-catalog-3","title":"codex-review-hook.py","projectId":"claude-code-autopilot","source":"claude-code-autopilot/.claude/hooks/codex-review-hook.py","date":"2026-09-17","excerpt":"Review integration implementation.","url":"/evidence/claude-code-autopilot-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"claude-code-autopilot-catalog-4","title":"claude-code-autopilot git history: 2c1f5ec, 19c6bf1, a6206c6","projectId":"claude-code-autopilot","source":"claude-code-autopilot git history: 2c1f5ec, 19c6bf1, a6206c6","date":"2026-09-17","excerpt":"Donald installer and review-workflow revisions.","url":"/evidence/claude-code-autopilot-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-foundation-catalog-1","title":"README.md","projectId":"zhengmind-foundation","source":"ZhengMind/README.md","date":"2026-09-17","excerpt":"Explicit local/mock boundary, canonical work model, implemented MVP surfaces, and future concepts.","url":"/evidence/zhengmind-foundation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-foundation-catalog-2","title":"service.py","projectId":"zhengmind-foundation","source":"ZhengMind/hermes/task_graph/service.py","date":"2026-09-17","excerpt":"Task-tree service and claiming implementation.","url":"/evidence/zhengmind-foundation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-foundation-catalog-3","title":"work_semantic_compatibility_contract.md","projectId":"zhengmind-foundation","source":"ZhengMind/docs/work_semantic_compatibility_contract.md","date":"2026-09-17","excerpt":"Work/task naming contract and compatibility limits.","url":"/evidence/zhengmind-foundation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-foundation-catalog-4","title":"architecture-review.md","projectId":"zhengmind-foundation","source":"ZhengMindYAGNI/docs/architecture-review.md","date":"2026-09-17","excerpt":"Concrete documented relationship between this predecessor and the simplification study.","url":"/evidence/zhengmind-foundation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zhengmind-foundation-catalog-5","title":"ZhengMind git history","projectId":"zhengmind-foundation","source":"ZhengMind git history","date":"2026-09-17","excerpt":"Clean upstream history under yq77zs73/Zheng Shen; no Donald identity equivalence established.","url":"/evidence/zhengmind-foundation-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"serena-foundation-catalog-1","title":"README.md","projectId":"serena-foundation","source":"serena/README.md","date":"2026-09-17","excerpt":"Upstream semantic tools, language-server design, and MCP/client integrations.","url":"/evidence/serena-foundation-catalog-1","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"serena-foundation-catalog-2","title":"serena git remote and history","projectId":"serena-foundation","source":"serena git remote and history","date":"2026-09-17","excerpt":"Upstream oraios/serena origin, upstream contributors, and clean tracked working tree.","url":"/evidence/serena-foundation-catalog-2","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-1","title":"README.md","projectId":"isafe-sop","source":"aisop/README.md","date":"2026-09-17","excerpt":"Product surfaces, module layout, mock default, optional real backends, and skipped-step acceptance demo.","url":"/evidence/isafe-sop-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-2","title":"fsm.py","projectId":"isafe-sop","source":"aisop/packages/aisop/sop_engine/fsm.py","date":"2026-09-17","excerpt":"DAG mandatory ancestors, three-frame persistence, skip/wrong-order reclassification, bounded normalized soft-DTW.","url":"/evidence/isafe-sop-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-3","title":"qwen_vl.py","projectId":"isafe-sop","source":"aisop/packages/aisop/adjudication/clients/qwen_vl.py","date":"2026-09-17","excerpt":"Lazy-loaded Qwen2.5-VL, NF4 quantization, bounded visual input and generation, uncertainty-preserving output parsing.","url":"/evidence/isafe-sop-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-4","title":"sqlite.py","projectId":"isafe-sop","source":"aisop/packages/aisop/storage/repos/sqlite.py","date":"2026-09-17","excerpt":"Audit hash creation and chain validation.","url":"/evidence/isafe-sop-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-5","title":"sop_compiler.py","projectId":"isafe-sop","source":"aisop/apps/api/sop_compiler.py","date":"2026-09-17","excerpt":"Deterministic compiler implementation and explicit LLM TODO.","url":"/evidence/isafe-sop-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"isafe-sop-catalog-6","title":"aisop git history: 62850bd, 0de3157, 4579edc","projectId":"isafe-sop","source":"aisop git history: 62850bd, 0de3157, 4579edc","date":"2026-09-17","excerpt":"Donald Shen commits for backend implementation, integration, and confirmed bug repairs.","url":"/evidence/isafe-sop-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-1","title":"CMakeLists.txt","projectId":"cpp-goroutine","source":"go-rountine/CMakeLists.txt","date":"2026-09-17","excerpt":"C++17 library and echo/timer/basic-test build targets.","url":"/evidence/cpp-goroutine-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-2","title":"scheduler.cpp","projectId":"cpp-goroutine","source":"go-rountine/src/scheduler.cpp","date":"2026-09-17","excerpt":"Context creation, worker scheduling and parking synchronization.","url":"/evidence/cpp-goroutine-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-3","title":"goroutine.cpp","projectId":"cpp-goroutine","source":"go-rountine/src/goroutine.cpp","date":"2026-09-17","excerpt":"mmap allocation with mprotect guard page.","url":"/evidence/cpp-goroutine-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-4","title":"io_scheduler.cpp","projectId":"cpp-goroutine","source":"go-rountine/src/io_scheduler.cpp","date":"2026-09-17","excerpt":"epoll/eventfd and timer heap.","url":"/evidence/cpp-goroutine-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-5","title":"hook.cpp","projectId":"cpp-goroutine","source":"go-rountine/src/hook.cpp","date":"2026-09-17","excerpt":"libc I/O and sleep interception with yielding behavior.","url":"/evidence/cpp-goroutine-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-6","title":"mutex.cpp","projectId":"cpp-goroutine","source":"go-rountine/src/mutex.cpp","date":"2026-09-17","excerpt":"Cooperative mutex and wait-group behavior.","url":"/evidence/cpp-goroutine-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cpp-goroutine-catalog-7","title":"test_echo.cpp","projectId":"cpp-goroutine","source":"go-rountine/examples/test_echo.cpp","date":"2026-09-17","excerpt":"A small end-to-end network example, not a stress benchmark.","url":"/evidence/cpp-goroutine-catalog-7","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"applypilot-adaptation-catalog-1","title":"llm.py","projectId":"applypilot-adaptation","source":"applypilot/src/applypilot/llm.py","date":"2026-09-17","excerpt":"Local ClaudeCLIClient and provider-selection changes versus upstream.","url":"/evidence/applypilot-adaptation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"applypilot-adaptation-catalog-2","title":"scorer.py","projectId":"applypilot-adaptation","source":"applypilot/src/applypilot/scoring/scorer.py","date":"2026-09-17","excerpt":"Configurable parallel scoring and incremental SQLite commits.","url":"/evidence/applypilot-adaptation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"applypilot-adaptation-catalog-3","title":"README.md","projectId":"applypilot-adaptation","source":"applypilot/README.md","date":"2026-09-17","excerpt":"Explicit upstream project attribution and pipeline scope.","url":"/evidence/applypilot-adaptation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"applypilot-adaptation-catalog-4","title":"applypilot working-tree diff","projectId":"applypilot-adaptation","source":"applypilot working-tree diff","date":"2026-09-17","excerpt":"Two locally modified source files; upstream history has no Donald-attributed commits.","url":"/evidence/applypilot-adaptation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"invoice-authentication-catalog-1","title":"README.md","projectId":"invoice-authentication","source":"dataannotation/my-project/README.md","date":"2026-09-17","excerpt":"Explicit upstream repository link and named original author.","url":"/evidence/invoice-authentication-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"invoice-authentication-catalog-2","title":"auth.service.ts","projectId":"invoice-authentication","source":"dataannotation/my-project/invoice-backend/src/auth/auth.service.ts","date":"2026-09-17","excerpt":"Local credential verification, token issuing and refresh rotation implementation.","url":"/evidence/invoice-authentication-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"invoice-authentication-catalog-3","title":"refreshTokenCookie.ts","projectId":"invoice-authentication","source":"dataannotation/my-project/invoice-backend/src/auth/refreshTokenCookie.ts","date":"2026-09-17","excerpt":"Cookie transport and parsing.","url":"/evidence/invoice-authentication-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"invoice-authentication-catalog-4","title":"authRoutes.test.ts","projectId":"invoice-authentication","source":"dataannotation/my-project/invoice-backend/tests/unit/auth/authRoutes.test.ts","date":"2026-09-17","excerpt":"Tests specify HttpOnly rotation and absence of refresh secrets from JSON.","url":"/evidence/invoice-authentication-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"invoice-authentication-catalog-5","title":"auth.service.ts","projectId":"invoice-authentication","source":"dataannotation/my-project copy/invoice-backend/src/auth/auth.service.ts","date":"2026-09-17","excerpt":"Alternative repository-based session/auth design; variants grouped.","url":"/evidence/invoice-authentication-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cuda-introduction-catalog-1","title":"README.md","projectId":"cuda-introduction","source":"uiuc/ECE408/lab0/README.md","date":"2026-09-17","excerpt":"Explicitly states lab0 code is complete and provided.","url":"/evidence/cuda-introduction-catalog-1","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cuda-introduction-catalog-2","title":"lab0.cu","projectId":"cuda-introduction","source":"uiuc/ECE408/lab0/lab0.cu","date":"2026-09-17","excerpt":"Supplied device-information program.","url":"/evidence/cuda-introduction-catalog-2","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"cuda-introduction-catalog-3","title":"README.md","projectId":"cuda-introduction","source":"uiuc/ECE408/README.md","date":"2026-09-17","excerpt":"Course repository context; Git commit 2945941 retrieves release content.","url":"/evidence/cuda-introduction-catalog-3","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"blockchain-principles-catalog-1","title":"slides","projectId":"blockchain-principles","source":"uiuc/blockchains_principles/slides","date":"2026-09-17","excerpt":"Lecture-slide collection inventoried by filename.","url":"/evidence/blockchain-principles-catalog-1","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"blockchain-principles-catalog-2","title":"notes","projectId":"blockchain-principles","source":"uiuc/blockchains_principles/notes","date":"2026-09-17","excerpt":"Course-note collection inventoried by filename.","url":"/evidence/blockchain-principles-catalog-2","context":"Source summary from the project inventory. This is an upstream reference, not original work by Donald. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"kalshibot-catalog-1","title":"refactor.md","projectId":"kalshibot","source":"kalshibot/docs/refactor.md","date":"2026-09-17","excerpt":"Independent snapshot provenance, responsibility decomposition and explicit equivalence limits.","url":"/evidence/kalshibot-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"kalshibot-catalog-2","title":"verification.md","projectId":"kalshibot","source":"kalshibot/docs/verification.md","date":"2026-09-17","excerpt":"Dated offline engineering validation and remaining real-account acceptance boundary.","url":"/evidence/kalshibot-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"kalshibot-catalog-3","title":"client.py","projectId":"kalshibot","source":"kalshibot/src/kalshibot/client.py","date":"2026-09-17","excerpt":"Ambiguous non-idempotent creation is not automatically retried.","url":"/evidence/kalshibot-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"kalshibot-catalog-4","title":"Contribution history","projectId":"kalshibot","source":"kalshibot-membrane-crosssection/.git","date":"2026-09-17","excerpt":"History includes breadth-first cross-sectional sweep; this is a related branch of membrane.","url":"/evidence/kalshibot-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"kalshibot-catalog-5","title":"Contribution history","projectId":"kalshibot","source":"kalshibot-membrane-sweepfreeze/.git","date":"2026-09-17","excerpt":"Historical membrane snapshot with paper-rig hardening history; group with parent.","url":"/evidence/kalshibot-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"statistical-arbitrage-catalog-1","title":"pairability.py","projectId":"statistical-arbitrage","source":"pair-trade/src/graph/pairability.py","date":"2026-09-17","excerpt":"Pairability graph, forward-correlation estimator and graph builder are implemented.","url":"/evidence/statistical-arbitrage-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"statistical-arbitrage-catalog-2","title":"net_alpha.py","projectId":"statistical-arbitrage","source":"pair-trade/src/alpha/net_alpha.py","date":"2026-09-17","excerpt":"ExpectedNetAlphaModel and score decomposition source.","url":"/evidence/statistical-arbitrage-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"statistical-arbitrage-catalog-3","title":"production_engine.py","projectId":"statistical-arbitrage","source":"pair-trade/src/pipeline/production_engine.py","date":"2026-09-17","excerpt":"Integration source exists; runtime deployment not verified.","url":"/evidence/statistical-arbitrage-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"statistical-arbitrage-catalog-4","title":"README.md","projectId":"statistical-arbitrage","source":"sky_discover_pair_trade/README.md","date":"2026-09-17","excerpt":"Six EVOLVE-BLOCK tasks and explicit separation of search harness, promotion gate and integrated validation.","url":"/evidence/statistical-arbitrage-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"options-variance-catalog-1","title":"README.md","projectId":"options-variance","source":"IV_trading/README.md","date":"2026-09-17","excerpt":"Source layout for pricing, chains, strategies, backtesting and execution.","url":"/evidence/options-variance-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"options-variance-catalog-2","title":"RESEARCH.md","projectId":"options-variance","source":"IV_trading/RESEARCH.md","date":"2026-09-17","excerpt":"Variance-risk-premium hypotheses and target/input timing distinction.","url":"/evidence/options-variance-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"options-variance-catalog-3","title":"README.md","projectId":"options-variance","source":"IV_trading/evc1/README.md","date":"2026-09-17","excerpt":"Earnings event-variance math, prior-only shrinkage and natural-quote structure evaluation.","url":"/evidence/options-variance-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"options-variance-catalog-4","title":"engine.py","projectId":"options-variance","source":"IV_trading/engine.py","date":"2026-09-17","excerpt":"LiveEngine source implements an order decision flow; not execution evidence.","url":"/evidence/options-variance-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-1","title":"PREREG.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/metalabel/PREREG.md","date":"2026-09-17","excerpt":"June 2026 preregistration separates meta-label selection from timing/sizing and specifies PIT availability, entry and evaluation gates. Goals are not outcomes.","url":"/evidence/lppls-research-cockpits-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-2","title":"serve.py","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/metalabel/dashboard/serve.py","date":"2026-09-17","excerpt":"Current FastAPI source includes startup reconciliation plus date, regime, signal, candidate, simulated-book, performance and status endpoints.","url":"/evidence/lppls-research-cockpits-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-3","title":"update_daily.py","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/metalabel/dashboard/update_daily.py","date":"2026-09-17","excerpt":"Current source includes snapshots, restoration, staged updates, frozen-window regression guards, dry verification and explicit status handling. Code presence does not establish that scheduling is active.","url":"/evidence/lppls-research-cockpits-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-4","title":"meta.py","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/metalabel/code/meta.py","date":"2026-09-17","excerpt":"Current source contains purged OOF fitting, isotonic calibration, ensemble construction and strict-OOS evaluation.","url":"/evidence/lppls-research-cockpits-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-5","title":"RUN.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/global_bubble/RUN.md","date":"2026-09-17","excerpt":"Separate global monitor for 23 indices/ETFs using trailing nested LPPLS fits; supported by lppls_scan.py and a cockpit API.","url":"/evidence/lppls-research-cockpits-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-6","title":"VERDICT.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/frame_attack/reports/VERDICT.md","date":"2026-09-17","excerpt":"Dated hypothesis scorecard records failed conditional-information tests, a confirmatory gate not run, null exploratory results and killed mechanisms.","url":"/evidence/lppls-research-cockpits-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-7","title":"VERDICT.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/tail_runway/reports/VERDICT.md","date":"2026-09-17","excerpt":"The proposed supply-runway primary hypothesis was rejected; a later candidate remained parked pending stronger forward evidence.","url":"/evidence/lppls-research-cockpits-catalog-7","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-8","title":"MECHANISM_VERDICT.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/pm_system/reports/MECHANISM_VERDICT.md","date":"2026-09-17","excerpt":"Mechanism gate distinguishes a fitted effect from an explained decomposition and quarantines the unexplained residual.","url":"/evidence/lppls-research-cockpits-catalog-8","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"lppls-research-cockpits-catalog-9","title":"research_plan.md","projectId":"lppls-research-cockpits","source":"A_Share/openassetpricing/report/cbsf_factor/research_plan.md","date":"2026-09-17","excerpt":"Cost-basis factor study explicitly notes that vendor chips are imputed from price/volume, so the experiment tests predictive transforms rather than disposition-physics claims.","url":"/evidence/lppls-research-cockpits-catalog-9","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zero-human-hedge-catalog-1","title":"README.md","projectId":"zero-human-hedge","source":"zerohumanhedge/README.md","date":"2026-09-17","excerpt":"Bootstrap and later research/backtest flow.","url":"/evidence/zero-human-hedge-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zero-human-hedge-catalog-2","title":"cli.py","projectId":"zero-human-hedge","source":"zerohumanhedge/src/zerohumanhedge/cli.py","date":"2026-09-17","excerpt":"run_research integrates data, features, optimizer and artifact exports; parity and Paperclip commands are implemented.","url":"/evidence/zero-human-hedge-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"zero-human-hedge-catalog-3","title":"backtest.py","projectId":"zero-human-hedge","source":"zerohumanhedge/apps/trading_node/backtest.py","date":"2026-09-17","excerpt":"Nautilus parity-backtest module referenced by the concrete CLI.","url":"/evidence/zero-human-hedge-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mention-market-model-catalog-1","title":"README.md","projectId":"mention-market-model","source":"prediction_market/kalshi_wc_mention_analysis/README.md","date":"2026-09-17","excerpt":"Separate original study, source coverage hierarchy, probabilistic model and ablation discussion.","url":"/evidence/mention-market-model-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mention-market-model-catalog-2","title":"model.py","projectId":"mention-market-model","source":"prediction_market/kalshi_wc_mention_analysis/corpus_engine/model.py","date":"2026-09-17","excerpt":"MentionModel implements word/context/crew shifts and corpus tilt.","url":"/evidence/mention-market-model-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mention-market-model-catalog-3","title":"eval_corpus_value.py","projectId":"mention-market-model","source":"prediction_market/kalshi_wc_mention_analysis/eval_corpus_value.py","date":"2026-09-17","excerpt":"Model-comparison entrypoint; results not rerun.","url":"/evidence/mention-market-model-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"factor-replication-catalog-1","title":"README.md","projectId":"factor-replication","source":"A_Share/openassetpricing/README.md","date":"2026-09-17","excerpt":"Upstream OpenSourceAP replication scope, universe rules and announcement-time convention.","url":"/evidence/factor-replication-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"factor-replication-catalog-2","title":"README.md","projectId":"factor-replication","source":"A_Share/factorResearch/paper1/README.md","date":"2026-09-17","excerpt":"Tushare factor calculation, processing, combination and index-enhancement evaluation pipeline.","url":"/evidence/factor-replication-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"factor-replication-catalog-3","title":"README.md","projectId":"factor-replication","source":"A_Share/factorResearch-research/paper1/README.md","date":"2026-09-17","excerpt":"Related copy with the same pipeline scope; avoid duplicate accomplishments.","url":"/evidence/factor-replication-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"factor-replication-catalog-4","title":"readme.md","projectId":"factor-replication","source":"A_Share/hikyuu/readme.md","date":"2026-09-17","excerpt":"Upstream framework documentation.","url":"/evidence/factor-replication-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"factor-replication-catalog-5","title":"pytdx_to_h5.py","projectId":"factor-replication","source":"A_Share/hikyuu/hikyuu/data/pytdx_to_h5.py","date":"2026-09-17","excerpt":"Only tracked local modification identified in Git status.","url":"/evidence/factor-replication-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mean-reversion-system-catalog-1","title":"2024-12-22-mean-reversion-design.md","projectId":"mean-reversion-system","source":"mean-reversion/docs/plans/2024-12-22-mean-reversion-design.md","date":"2026-09-17","excerpt":"Design explicitly separates EOD signals, next-open trades, hedging and market frictions.","url":"/evidence/mean-reversion-system-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mean-reversion-system-catalog-2","title":"CH3_IMPLEMENTATION_SUMMARY.md","projectId":"mean-reversion-system","source":"mean-reversion/CH3_IMPLEMENTATION_SUMMARY.md","date":"2026-09-17","excerpt":"Factor-alpha evaluation implementation; comparison benchmark is not a project outcome.","url":"/evidence/mean-reversion-system-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"mean-reversion-system-catalog-3","title":"Contribution history","projectId":"mean-reversion-system","source":"mean-reversion/.git","date":"2026-09-17","excerpt":"December 2025 Donald-attributed source and verification changes.","url":"/evidence/mean-reversion-system-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-signal-ingestion-catalog-1","title":"README.md","projectId":"market-signal-ingestion","source":"poly/README.md","date":"2026-09-17","excerpt":"Defines ingestion scope and explicitly excludes downstream trading.","url":"/evidence/market-signal-ingestion-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-signal-ingestion-catalog-2","title":"replay.py","projectId":"market-signal-ingestion","source":"poly/replay.py","date":"2026-09-17","excerpt":"Replay entrypoint exists.","url":"/evidence/market-signal-ingestion-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-signal-ingestion-catalog-3","title":"Contribution history","projectId":"market-signal-ingestion","source":"poly/.git","date":"2026-09-17","excerpt":"394c517 records dashboard; bec372a records replay and alert sinks.","url":"/evidence/market-signal-ingestion-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"technical-indicator-evaluation-catalog-1","title":"README.md","projectId":"technical-indicator-evaluation","source":"indicator_test/README.md","date":"2026-09-17","excerpt":"Universal scalar indicator contract and multi-fold evaluation interface.","url":"/evidence/technical-indicator-evaluation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"technical-indicator-evaluation-catalog-2","title":"evaluate.py","projectId":"technical-indicator-evaluation","source":"indicator_test/evaluate.py","date":"2026-09-17","excerpt":"Fold planning, timeout handling, failure replay, IC computation and aggregation functions inspected via AST.","url":"/evidence/technical-indicator-evaluation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-1","title":"README.md","projectId":"market-mechanism-studies","source":"A_Share/SW_research/README.md","date":"2026-09-17","excerpt":"Frozen episode study with per-mechanism status and corrected/retracted conclusions.","url":"/evidence/market-mechanism-studies-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-2","title":"REPORT.md","projectId":"market-mechanism-studies","source":"A_Share/hm_study/REPORT.md","date":"2026-09-17","excerpt":"Documents negative results and execution limitations for disclosure-following hypotheses.","url":"/evidence/market-mechanism-studies-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-3","title":"RESULTS.md","projectId":"market-mechanism-studies","source":"A_Share/ipo_study/RESULTS.md","date":"2026-09-17","excerpt":"Empirical listing-day study; allocation probabilities and market eras are distinct.","url":"/evidence/market-mechanism-studies-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-4","title":"pull_dragon_tiger.py","projectId":"market-mechanism-studies","source":"alpha_testing/jump_excitation/pull_dragon_tiger.py","date":"2026-09-17","excerpt":"Functions cover disclosure pulls, down-day events, seat composition and forward-return analysis; inspected via AST.","url":"/evidence/market-mechanism-studies-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-5","title":"README.md","projectId":"market-mechanism-studies","source":"china_national_team_tracker/README.md","date":"2026-09-17","excerpt":"Defines observable proxy components, missing-feed handling and demo boundaries.","url":"/evidence/market-mechanism-studies-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"market-mechanism-studies-catalog-6","title":"engine.py","projectId":"market-mechanism-studies","source":"china_national_team_tracker/national_team_engine/engine.py","date":"2026-09-17","excerpt":"NationalTeamEngine and EngineResult source inspected via AST.","url":"/evidence/market-mechanism-studies-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"finance-reading-and-falsification-catalog-1","title":"research.md","projectId":"finance-reading-and-falsification","source":"strategy_momentum/research.md","date":"2026-09-17","excerpt":"Comparative survey of momentum algorithms; figures are literature claims.","url":"/evidence/finance-reading-and-falsification-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"finance-reading-and-falsification-catalog-2","title":"Brooks_AShare_System.md","projectId":"finance-reading-and-falsification","source":"A_Share/al_brooks/Brooks_AShare_System.md","date":"2026-09-17","excerpt":"Explicit report/blueprint scope, no code claim.","url":"/evidence/finance-reading-and-falsification-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"finance-reading-and-falsification-catalog-3","title":"assessment.md","projectId":"finance-reading-and-falsification","source":"A_Share/yi/docs/assessment.md","date":"2026-09-17","excerpt":"Critical assessment of a book’s predictive claims; preserve as study synthesis, not personal quotation.","url":"/evidence/finance-reading-and-falsification-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"finance-reading-and-falsification-catalog-4","title":"test_plan.md","projectId":"finance-reading-and-falsification","source":"A_Share/insider_case_study/test_plan.md","date":"2026-09-17","excerpt":"Maps informal trading claims to testability and data requirements.","url":"/evidence/finance-reading-and-falsification-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"csc-tactics-catalog-1","title":"README.md","projectId":"csc-tactics","source":"soccer/README.md","date":"2026-09-17","excerpt":"Content counts, architecture, Canvas fallback and deployment URL.","url":"/evidence/csc-tactics-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source","sourceUrl":"https://csc.donald7.us/"},{"id":"csc-tactics-catalog-2","title":"4231_测试与限制.md","projectId":"csc-tactics","source":"soccer/docs/4231_测试与限制.md","date":"2026-09-17","excerpt":"Two-ball-to-one-ball lifecycle bug and dated validation boundaries.","url":"/evidence/csc-tactics-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source","sourceUrl":"https://csc.donald7.us/"},{"id":"csc-tactics-catalog-3","title":"drills.json","projectId":"csc-tactics","source":"soccer/src/data/drills.json","date":"2026-09-17","excerpt":"Actual structured training segments, including reflection prompts.","url":"/evidence/csc-tactics-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source","sourceUrl":"https://csc.donald7.us/"},{"id":"csc-tactics-catalog-4","title":"4231_研究与设计说明.md","projectId":"csc-tactics","source":"soccer/docs/4231_研究与设计说明.md","date":"2026-09-17","excerpt":"Distinguishes research support from design choices and unproven outcomes.","url":"/evidence/csc-tactics-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source","sourceUrl":"https://csc.donald7.us/"},{"id":"youtube-knowledge-catalog-1","title":"pipeline.py","projectId":"youtube-knowledge","source":"youtube_knowledge/pipeline.py","date":"2026-09-17","excerpt":"Two-stage processing, subprocess model invocation, bounded concurrency, per-slug locking, atomic writes.","url":"/evidence/youtube-knowledge-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"youtube-knowledge-catalog-2","title":"server.py","projectId":"youtube-knowledge","source":"youtube_knowledge/server.py","date":"2026-09-17","excerpt":"NDJSON streaming with heartbeat, channel and feed endpoints.","url":"/evidence/youtube-knowledge-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"youtube-knowledge-catalog-3","title":"distill.py","projectId":"youtube-knowledge","source":"youtube_knowledge/distill.py","date":"2026-09-17","excerpt":"Source-linked distillation and best-effort knowledge graph.","url":"/evidence/youtube-knowledge-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"youtube-knowledge-catalog-4","title":"graph.py","projectId":"youtube-knowledge","source":"youtube_knowledge/graph.py","date":"2026-09-17","excerpt":"Allowed source-ID checks and cross-video merge requirement.","url":"/evidence/youtube-knowledge-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"youtube-knowledge-catalog-5","title":"test_pipeline.py","projectId":"youtube-knowledge","source":"youtube_knowledge/test_pipeline.py","date":"2026-09-17","excerpt":"Deterministic tests for model routing, missing captions, event ordering, and concurrent profile updates.","url":"/evidence/youtube-knowledge-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"youtube-knowledge-catalog-6","title":"youtube_knowledge git history: 7b4b814, 8c3b65a, e72b179","projectId":"youtube-knowledge","source":"youtube_knowledge git history: 7b4b814, 8c3b65a, e72b179","date":"2026-09-17","excerpt":"Donald-attributed streaming/channel/feed implementation.","url":"/evidence/youtube-knowledge-catalog-6","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"piano-midi-catalog-1","title":"pyproject.toml","projectId":"piano-midi","source":"pianovision_midi_converter/pyproject.toml","date":"2026-09-17","excerpt":"Dependencies and declared purpose; Git commit e2b2554 by donald7.","url":"/evidence/piano-midi-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"piano-midi-catalog-2","title":"transcriber.py","projectId":"piano-midi","source":"pianovision_midi_converter/pipeline/transcriber.py","date":"2026-09-17","excerpt":"Upstream transcription call and CUDA-to-CPU fallback.","url":"/evidence/piano-midi-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"piano-midi-catalog-3","title":"postprocessor.py","projectId":"piano-midi","source":"pianovision_midi_converter/pipeline/postprocessor.py","date":"2026-09-17","excerpt":"Timing, velocity and overlap processing.","url":"/evidence/piano-midi-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"piano-midi-catalog-4","title":"tasks.py","projectId":"piano-midi","source":"pianovision_midi_converter/tasks.py","date":"2026-09-17","excerpt":"Queue, state transitions, deduplication and saved library.","url":"/evidence/piano-midi-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"browser-repl-catalog-1","title":"package.json","projectId":"browser-repl","source":"dataannotation/may22/package.json","date":"2026-09-17","excerpt":"Vue/Vite, CodeMirror, JavaScript/Python language tools.","url":"/evidence/browser-repl-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"browser-repl-catalog-2","title":"runners.js","projectId":"browser-repl","source":"dataannotation/may22/src/lib/runners.js","date":"2026-09-17","excerpt":"Iframe JavaScript execution and lazy Pyodide Python implementation.","url":"/evidence/browser-repl-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"browser-repl-catalog-3","title":"storage.js","projectId":"browser-repl","source":"dataannotation/may22/src/lib/storage.js","date":"2026-09-17","excerpt":"Local file persistence.","url":"/evidence/browser-repl-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"browser-repl-catalog-4","title":"App.vue","projectId":"browser-repl","source":"dataannotation/may22codex/src/App.vue","date":"2026-09-17","excerpt":"Worker-based execution, termination timeouts, save/export and formatting.","url":"/evidence/browser-repl-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"daily-news-terminal-catalog-1","title":"README.md","projectId":"daily-news-terminal","source":"daily-news-terminal-docker/README.md","date":"2026-09-17","excerpt":"Feature inventory, public-source integrations, no seeded fallback, and read-only trading boundary.","url":"/evidence/daily-news-terminal-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"daily-news-terminal-catalog-2","title":"server.js","projectId":"daily-news-terminal","source":"daily-news-terminal-docker/server.js","date":"2026-09-17","excerpt":"Normalization, heuristic matching, timeouts/cache, source-quality metadata, and service implementation.","url":"/evidence/daily-news-terminal-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"daily-news-terminal-catalog-3","title":"app.js","projectId":"daily-news-terminal","source":"daily-news-terminal-docker/public/app.js","date":"2026-09-17","excerpt":"Interactive display implementation.","url":"/evidence/daily-news-terminal-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"daily-news-terminal-catalog-4","title":"package.json","projectId":"daily-news-terminal","source":"daily-news-terminal-docker/package.json","date":"2026-09-17","excerpt":"Node entrypoint and syntax-check scripts without declared third-party packages.","url":"/evidence/daily-news-terminal-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"video-transcriber-adaptation-catalog-1","title":"README.md","projectId":"video-transcriber-adaptation","source":"video_transribe/AI-Video-Transcriber/README.md","date":"2026-09-17","excerpt":"Identifies upstream repository and original subtitle-first architecture.","url":"/evidence/video-transcriber-adaptation-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"video-transcriber-adaptation-catalog-2","title":"claude_client.py","projectId":"video-transcriber-adaptation","source":"video_transribe/AI-Video-Transcriber/backend/claude_client.py","date":"2026-09-17","excerpt":"Local async Claude Agent SDK adapter.","url":"/evidence/video-transcriber-adaptation-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"video-transcriber-adaptation-catalog-3","title":"transcriber.py","projectId":"video-transcriber-adaptation","source":"video_transribe/AI-Video-Transcriber/backend/transcriber.py","date":"2026-09-17","excerpt":"Working-tree diff changes CPU int8 to CUDA float16.","url":"/evidence/video-transcriber-adaptation-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"video-transcriber-adaptation-catalog-4","title":"video_processor.py","projectId":"video-transcriber-adaptation","source":"video_transribe/AI-Video-Transcriber/backend/video_processor.py","date":"2026-09-17","excerpt":"Local platform retrieval modifications.","url":"/evidence/video-transcriber-adaptation-catalog-4","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"video-transcriber-adaptation-catalog-5","title":"smoke_test.py","projectId":"video-transcriber-adaptation","source":"video_transribe/AI-Video-Transcriber/backend/smoke_test.py","date":"2026-09-17","excerpt":"Local smoke-test script; not executed in this audit.","url":"/evidence/video-transcriber-adaptation-catalog-5","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"personal-website-catalog-1","title":"knowledge.ts","projectId":"personal-website","source":"personal-website/content/knowledge.ts","date":"2026-09-17","excerpt":"Shared candidate, project, claim and evidence model.","url":"/evidence/personal-website-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"personal-website-catalog-2","title":"guide.ts","projectId":"personal-website","source":"personal-website/lib/guide.ts","date":"2026-09-17","excerpt":"Local Qwen selects existing reviewed claims and falls back transparently when unavailable.","url":"/evidence/personal-website-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"personal-website-catalog-3","title":"server.ts","projectId":"personal-website","source":"personal-website/lib/server.ts","date":"2026-09-17","excerpt":"Signed answer sharing and immutable evidence snapshots.","url":"/evidence/personal-website-catalog-3","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"technology-society-writing-catalog-1","title":"american_graffiti_essay.tex","projectId":"technology-society-writing","source":"uiuc/hist/american_graffiti_essay.tex","date":"2026-09-17","excerpt":"Donald-named source document with substantive analysis of transport technology, identity and social change.","url":"/evidence/technology-society-writing-catalog-1","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"technology-society-writing-catalog-2","title":"fordlandia_essay.pdf","projectId":"technology-society-writing","source":"uiuc/hist/fordlandia_essay.pdf","date":"2026-09-17","excerpt":"Related essay artifact observed by filename; content not audited.","url":"/evidence/technology-society-writing-catalog-2","context":"Source summary from the project inventory. Historical results are identified in the project notes; source inspection is not a new benchmark.","kind":"source"},{"id":"illinix-391-catalog-1","title":"Donald’s project record · Spring 2025","projectId":"illinix-391","source":"Donald’s project record · Spring 2025","date":"2026-09-17","excerpt":"Illinix 391 — Unix-like OS Kernel. Virtual memory, fork/exec, 15+ syscalls, VIRTIO, KTFS, ELF loader, preemptive multitasking, pipes, and shell redirection. Champaign, Illinois.","url":"/evidence/illinix-391-catalog-1","context":"Project scope and outcomes supplied directly by Donald. See the project scope note for the reporting period.","kind":"candidate"},{"id":"reaction-wheel-pendulum-catalog-1","title":"Donald’s project record · Fall 2024","projectId":"reaction-wheel-pendulum","source":"Donald’s project record · Fall 2024","date":"2026-09-17","excerpt":"Stable equilibrium control of a reaction-wheel pendulum using Lagrangian dynamics, three-state feedback, a Luenberger observer, and real-time friction compensation through Wincon. Champaign, Illinois.","url":"/evidence/reaction-wheel-pendulum-catalog-1","context":"Project scope and outcomes supplied directly by Donald. See the project scope note for the reporting period.","kind":"candidate"},{"id":"twentyfourpoints-catalog-1","title":"Donald’s project record · Summer 2025","projectId":"twentyfourpoints","source":"Donald’s project record · Summer 2025","date":"2026-09-17","excerpt":"Production-ready real-time multiplayer math game launched in 72 hours, with 760+ unique puzzles solved and reported 100% uptime. Custom validation, auto-balancing, tests, migrations, authentication, and Supabase ELO rankings.","url":"/evidence/twentyfourpoints-catalog-1","context":"Project scope and outcomes supplied directly by Donald. See the project scope note for the reporting period.","kind":"candidate","sourceUrl":"https://twentyfourpoints.com"}],"timeline":[{"date":"NOV 2025","title":"What does a model remember?","body":"Cross-attention, observed outcomes, and a paper reproduction.","project":"xtrend"},{"date":"DEC 2025","title":"What happens when paths collide?","body":"Five robots and 120 assignments in a fixed simulation study.","project":"swarm-navigation"},{"date":"MAR 2026","title":"What is the objective really optimizing?","body":"A forecast loss meets the exit rule that consumes it.","project":"chronos-knn"},{"date":"SEP 2026","title":"What happens between the happy paths?","body":"Order reconciliation and the lifecycle of an interactive scene.","project":"kalshibot"}],"connections":[{"label":"ROBOTICS × EMBEDDED VISION","title":"Seeing is only the beginning.","body":"A lane mask becomes a vehicle-control input. A face landmark becomes part of a time-dependent alert. In both projects, the interesting engineering is how perception turns into a decision—and what happens when that input is unreliable.","projects":["autonomous-driving-lab","driver-monitoring"],"kind":"SUPPORTED INTERPRETATION"},{"label":"AUTOMATION × AGENT SYSTEMS","title":"Who is allowed to take the next step?","body":"ISAFE checks whether assembly steps satisfy their prerequisites. ZhengMindYAGNI makes agent work follow an explicit state machine. The shared problem is making progress depend on something more precise than an optimistic status label.","projects":["isafe-sop","zhengmind-yagni"],"kind":"SUPPORTED INTERPRETATION"},{"label":"MODELS × INTERACTIVE TOOLS","title":"The work around the model.","body":"A transcription model produces notes; a usable piano tool also needs timing cleanup and recovery paths. Local inference similarly needs budgets, structured outputs and failure handling. Integration is a substantial part of both projects.","projects":["piano-midi","local-qwen-inference"],"kind":"SUPPORTED INTERPRETATION"}],"voiceCorpus":[],"limits":"Unknown fields are intentionally null or absent. Academic topics describe coursework, not professional proficiency. Experience and project outcomes identify candidate-provided sources; publication bibliography is linked to DOI records. Do not infer beliefs, adoption, or profitability from project presence."}