Results for “recursive-self-improvement”

Stateful single-mission improvement loop with strict evaluator contract, markdown decision logs, and max-runtime stop behavior