My research work — papers, active projects, and the code behind them. Everything is open source and reproducible; each entry links to its full materials.
Preprint — 22-page paper with proofs and 11 reproducible simulations
A coupled-ODE model of when competitive recursive self-improvement collapses into single-agent dominance. The theorems are conditional, and calibration against Epoch AI compute and benchmark data does not yet support the critical divergence assumption — the paper says so.
View PDF Code
Open-source instrument — 183 GitHub stars
Applies a Jacobian activation lens at nine depths inside a small LLM during live conversation, decoding hidden states through the vocabulary — with per-token inspection and session replay that runs in a browser without a GPU.
Live demo Code
Computational study · undergraduate research, UNC Greensboro
Simulations of emergent flocking, order–disorder phase transitions, predator–prey dynamics, and self-organized criticality in sandpile models — self-organization in systems of many simple agents.
Code