Skip to content
Live
Trending storyPast story

Anthropic's collaborative proof of Fermat's Last Theorem

2 reports2 outlets5 days agoUpdate

The story so far

AI overview

Anthropic announced a toy experiment: 1000 agents running Claude Fable 5.2 early versions, organized and assigned tasks solely using Git and a shared Postgres database in a network sandbox, completed the formal proof of Fermat's Last Theorem in 28.7 hours, and did not use scaffolding or researcher guidance provided by Prove2Me. For security, each agent worked in a network sandbox; the classifier checked every shell command and file edit before execution, and the offline monitor reviewed team actions. This run only enabled action-based monitors; parallel CoT monitors were added later. Anthropic stated that the output tokens were a third of the September version, and the proof length was a seventh of it.

AI-written from the reports · Updated 5 days ago

Coverage timeline

Follow the coverage to see every side of the story.

10 Nov
  1. Hacker News · 300+
    Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2 large models

    Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2, significantly improving in coding, reasoning, and visual tasks, supporting various complex applications.

9 Nov
  1. Claude.dev BlogFeatured
    Anthropic uses Claude Fable 5.2 to enable 1000 agents to safely collaborate on proving Fermat's Last Theorem

    Anthropic announces a toy experiment: 1000 agents running early versions of Claude Fable 5.2 in a network sandbox, organizing and dividing tasks solely with Git and a shared Postgres database, completing the formal proof of Fermat's Last Theorem in 28.7 hours, outputting tokens one-third of the September version and proof length one-seventh.

Heat over time

Now
3
Peak
810 Nov, 05:00
24h change
-49%

Hover or tap the chart to see each hour; use the arrow keys on a keyboard.

Related stories