Sources: Anthropic is expected to make its IPO prospectus public late September and complete the listing days before the US midterm elections in November (Echo Wang/Reuters)

Fermat's Last Theorem (FLT) in the Lean proof assistant is currently being worked on by a community effort, not attributed to Claude. Formal verification of mathematical arguments in Lean can reveal issues that are often overlooked. Claude's work on Lean formalization for a result related to the Riemann zeta function demonstrates progress, but does not mean Claude has completed the full proof of FLT.

The central public effort for FLT is the Imperial College London repository, which is an ongoing Lean formalization of the theorem and led by Kevin Buzzard. FLT states that there are no positive integer solutions to the equation (a^n + b^n = c^n) for integers (n ≥ 2). A 2025 arXiv paper reported a complete Lean formalization of FLT for regular primes, but this is a formalization of a special case, not the general theorem.

The current characterization of the Imperial College London FLT repository is that it's an ongoing community project formalizing FLT in Lean. Best et al. 2025 paper on FLT for regular primes is a complete formalization of a special case. Claude's mathematics formalization work on a related task to the Riemann zeta function is an example of progress in a difficult area, not evidence that Claude has formally verified the full proof of FLT.

The wait for FLT's full formalization stems from the fact that proof assistants like Lean require every definition, lemma, inference, and dependency to be encoded explicitly and checked by the system. Professionals working on such theorems often need to formalize supporting theory, reconcile notation, prove auxiliary results, and manage the code for review and maintenance.

While AI assistance can help with drafting, identifying lemmas, translating parts of an argument, and debugging, it is not enough to claim that the AI has completed the formal proof. Claude's work on a Lean formalization related to the Riemann zeta problem shows that modern language models can contribute to formal reasoning, but it should not be confused with a guarantee that Claude has independently formalized FLT or that the community project is finished.

Formal verification is crucial in scenarios where errors cannot be tolerated, such as software assurance, system safety, and critical business logic. While AI-assisted reasoning tools can help lower the effort of formal verification, they do not replace the need for expert review and validation of the results. Businesses should distinguish between what has been formally verified, what parts of a workflow are suitable for automation, and what still requires expert review.