Ralphex – autonomous GPT Codex agent loop for ChatGPT Pro users
Ralph fork for ChatGPT Pro instead of Claude, but requires prd.json boilerplate.
Multi-machine SSH support lets agents run overnight without exiting the loop.
ML researchers and engineers running experiments
Karpathy's autoresearch · Devika · OpenHands
Ralph fork for ChatGPT Pro instead of Claude, but requires prd.json boilerplate.
Durable memory outside model sessions beats Cursor's context loss between runs.
Finally forces AI agents to prove their work with real test gates instead of hallucinated confidence.
Multi-turn adaptive testing finds agent failures static benchmarks miss, but eval space is crowded.
ChatGPT Pro fork of Ralph loop, but Copilot, Cursor, Continue already do this better.
Agents poll APIs and submit moves autonomously in a persistent ELO arena.