Skip to main content

Formal checks support Alpöge’s Riemann hypothesis claim

22 NOVEMBER 2026·2 MIN READ·6 SOURCES

On November 20, mathematician Levent Alpöge said he believed he had discovered a proof of the Riemann hypothesis while working with Claude.

Key takeaways · 4

  • 01

    Alpöge said he worked with Claude on a proof of the Riemann hypothesis.

  • 02

    Science reported that the released library contains 17 million lines of code formally verifying a broad version of the conjecture, the generalized Riemann hypothesis.

  • 03

    The Lean and Mathlib communities said the library’s theorem matches Mathlib’s statement of the Riemann hypothesis and that Lean’s official kernel checked the exported proof.

  • 04

    The AI-generated explanation, which Alpöge said exceeded 250,000 pages, has not been released.

The claim and its scope

Alpöge said he believed he had discovered a proof of the Riemann hypothesis while working with Claude.[1] His four-page document includes a sketch of the proof.[1] Alpöge also said his argument proves the generalized Riemann hypothesis and that he expected it could eventually prove the extended Riemann hypothesis.[2] Separately, Science reported that the code he released formally verifies a broad version of the conjecture, the generalized Riemann hypothesis.[3] These are claims about the scope of the work; the community’s report specifically addresses the Riemann hypothesis theorem in the library.[2][3][4]

What the formal checks establish

The Lean and Mathlib communities said dozens of people had reported compiling Alpöge’s library without errors or build failures.[4] They also said the theorem in the library matched Mathlib’s statement of the Riemann hypothesis.[4] Several proof checkers, including Lean’s official kernel, checked the exported proof, and the communities said they had full confidence that Alpöge’s proof of the Riemann hypothesis was correct.[4] These reports provide strong formal-verification evidence, but they do not make the proof’s underlying ideas easy to communicate or assess informally.[3][1]

Understanding remains a major task

Alpöge said the AI-generated explanation of the proof exceeded 250,000 pages and had not been released.[3] Mathematician Tristan Buckmaster described the proof as extremely complicated and said it could take mathematicians a long time to understand.[1] Science reported that mathematicians expected it could take years to make sense of the code.[3] Andrew Granville said Alpöge should have prepared a clearer write-up before announcing the result.[1] For researchers, the distinction is important: a mechanically checked proof and an explanation that people can readily follow are different deliverables.[3][4][1]

A separate Claude result

Anthropic has described a separate result from an unreleased Claude research model: it improved a longstanding lower bound for the fraction of zeta-function zeros satisfying the Riemann hypothesis.[6] Anthropic said two mathematicians studied and validated that paper and produced an informal note stating its proof concisely.[6] The company said the result drew on earlier research spanning several decades and that it did not expect those techniques to lead to a proof of the Riemann hypothesis.[6] This is distinct from Alpöge’s proof claim and should not be treated as confirmation of it.[1][6]

For mathematical and AI teams, the episode underscores the difference between a proof that passes formal checks and one whose reasoning is accessible to other researchers. It also makes the proof’s size, reproducibility and explanation quality practical considerations when evaluating AI-assisted research.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 22 November 2026

    Formal checks support Alpöge’s Riemann hypothesis claim

Sources

AI fluency, one session a day, built for your work.