Hacker News · AI· bakigul·· 8 hours agoSelectedAI score77
Anthropic Experiment Demonstrates 1,000 Claude Agents Collaborating to Formalize Proof of Fermat's Last Theorem
Safely scaling multi-agent teams with Claude Fable 5.2nd PostgreSQL
AI guide
Anthropic ran a multi-agent experiment involving 1,000 early-version Claude Fable 5.2 agents, which autonomously formalized the proof of Fermat's Last Theorem in 28.7 hours without human intervention or specialized collaboration platforms.
Why it matters
The experiment demonstrated a mechanism where agent swarms achieve self-organized division of labor relying solely on general-purpose databases, offering a blueprint for sandboxing and monitoring long-horizon tasks.
Source:Hacker News · AI · claude.dev