Anthropic AI Formalizes Fermat's Last Theorem in 13 Million Lines
What happened: Anthropic announced that its Claude model produced the first end-to-end, machine-checked Lean 4 proof of Fermat's Last Theorem, generating 13 million lines of code over 11 days.
What happened: Anthropic announced that its Claude model produced the first end-to-end, machine-checked Lean 4 proof of Fermat's Last Theorem, generating 13 million lines of code over 11 days. The effort, led by Tianyi Peng (Columbia) and reviewed by mathematician Kevin Buzzard, leveraged the Prove2Me tool to coordinate dozens of parallel agents, ultimately proving over 30,000 theorems in the process.
Why it matters: This achievement marks a milestone in AI-driven mathematical formalization, demonstrating the ability to autonomously verify one of mathematics' most famous results. The scale—five times larger than the entire Mathlib library—highlights both the promise and the challenges of machine-checked proofs. It also raises questions about the future role of AI in mathematical research and the potential for new discoveries through automated reasoning.
Source: Anthropic, SiliconANGLE, TechTimes