Overview
- Levent Alpöge, an Anthropic researcher, published a four-page note and a Lean 4 code archive claiming that he and the company’s model Claude produced a formal proof of the Riemann Hypothesis and its generalized form, with Alpöge saying the full human-readable writeup would exceed about 250,000 pages.
- Alpöge’s release on Friday, Nov. 20 included a multipart archive that independent teams confirmed matches his files and began compiling into Mathlib, the Lean community library, but the automated compilation and machine checks are still in progress.
- Reviewers say three specific human checks remain before acceptance: confirm the formal statement exactly matches the standard Riemann Hypothesis, verify every formal definition matches standard mathematics, and ensure no hidden assumptions or unproven axioms appear in the code.
- Lean and Mathlib let a computer check each logical step, which can speed verification but cannot by itself replace expert peer review and community scrutiny needed for publication and prize eligibility under Clay Institute rules.
- The claim follows recent high-profile AI-linked mathematical announcements and raises practical questions about attribution, reproducibility, the labor of human verification, and how the math community will handle proofs too large for conventional exposition.