Anthropic published a blog post and GitHub repository on Saturday documenting Claude's formal verification of Fermat's Last Theorem in Lean 4, the interactive theorem prover. Fermat's Last Theorem — that no three positive integers a, b, c satisfy the equation aⁿ + bⁿ = cⁿ for any integer n greater than 2 — was conjectured by Pierre de Fermat in 1637 and first proven by Andrew Wiles in 1995 after 358 years of attempts. Wiles's proof is approximately 130 pages of dense mathematics. A Lean 4 formal verification is a mechanically checkable proof in which every logical step is expressed in a language that a computer can verify for consistency — a different kind of achievement than producing a human-readable argument, and in some ways a more demanding one. The Hacker News discussion, which reached 573 points by Saturday afternoon, concentrated on the precise meaning of the claim. Claude produced the Lean 4 proof structures; Lean 4's type-checking engine verified their logical consistency. Whether this constitutes AI-generated mathematics or AI-assisted formalisation of known mathematics is a distinction the AI reasoning community will debate. The practical significance is clearer: a language model producing mechanically verifiable proofs in a formal system is a capability milestone that has direct implications for automated theorem proving, mathematical research assistance, and formal verification of software systems.
LLMs
Google Releases Gemini 4 Argon — Tops Frontier Benchmarks and Targets Autonomous Cybersecurity
30 September: Google's Gemini 4 Argon leads the Vals AI benchmark index over OpenAI GPT-6 Astra and …