Open-source autoresearch turns coding agents into autonomous improvement loops (6k stars)

tom_doerr · x · 2026-09-04

autoresearch (6k stars, inspired by Karpathy's autoresearch) turns Claude Code, OpenCode, or OpenAI Codex into a relentless improvement engine: set a GOAL and the agent runs a plan → modify → verify → keep/discard loop against a mechanical metric, compounding gains overnight. "You don't need AGI. You need a goal, a metric, and a loop that never quits." v2.2.0; hook guardrails are Claude Code-only.

Original post →

More from coding & agent

coding & agent channel →