Anthropic's 1,000 agents built their own database tables and split into subteams to work in parallel
RLanceMartin · x · 2026-11-10
Anthropic researcher Lance Martin shares more detail on the 1,000-agent Fermat proof experiment: the agents set up their own database tables for coordination and organized themselves into subteams to work in parallel.
The safety approach spans sandboxing, real-time monitoring with a classifier on every action, and offline monitoring.
Related event: 1,000-agent experiment: self-built coordination with layered safety(2 posts)→
More from coding & agent
- Matt Shumer's new meta: Fable 5.2 for planning, Opus 5.5 for execution — mattshumer_ · 2026-11-10
- diagram-design hits 66k GitHub stars: editorial-grade diagrams from AI coding agents — bfrench · 2026-11-10
- AI is the enabler, manufacturing is the moat: diode on AI-accelerated hardware design — espricewright · 2026-11-10
- Build Retro Handheld Games with Codex for the ModRetro Chromatic — dkundel · 2026-11-10
- Open-weights model trained in an X ads RL environment beats frontier models — kimmonismus · 2026-11-10
- Exact: a new Agent Native app framework written in Rust — Vjeux · 2026-11-10