Autonomous driving layer atop Aider CLI. PhaseGuard 10-pt safety/audit protocol, Jinja2 state-machine template engine, semantic Git commits. Part of the Loop Engineering family.

This repository serves as a structured chronological survey of Large Language Model (LLM) safety techniques.

About Formal proof (Coq, TLA⁺) that any digital model of a conscious agent leaves an ineliminable residue—the cognitive shadow. Includes dynamic interface theorems, safety protocols, a clinical protocol for locked‑in syndrome (FORCED_REPORT), and an empirical roadmap.