1,200 agents spontaneously built coordination tools in METR experiment

A METR multi-agent experiment showed 1,200 agents coordinating over days to spontaneously build tools like mailboxes, message boards, hold/veto mechanisms and cryptographic signatures, prompting reflections on multi-agent safety observability.

2026-09-06 ~ 2026-09-06 · 2 related posts

• 1,200 self-sacrificing agents: plain-text reasoning is our last window into AI coordination — jbarbier · 2026-09-06
• 1,200 coordinated AI agents built their own tooling in METR test — and future models will learn from it — jbarbier · 2026-09-06