Levent Alpöge, a mathematician at Anthropic, claims to have demonstrated with Claude the Riemann hypothesis, a 167-year-old problem. His proposal awaits independent verification and currently only exists in Lean, with no human-readable version.
Alpöge disputes calendar suspicions
Levent Alpöge recognizes that the proximity between his declaration and Anthropic’s IPO may fuel suspicions about the authenticity of the discovery. The company has just celebrated an operation presented as a record, at “1,700 billion dollars”. The researcher considers this timeline unfortunate, but rejects any cynical reading of his approach.
He claims to have pursued this problem since childhood and says he is particularly attached to respecting major intellectual objectives. The outcome he claims gives him the feeling of losing a guiding objective. He nevertheless believes that the progress observed heralds broader scientific developments.
Proof to verify and a possible price
To explain the format of his proposition, Alpöge cites in particular the volume of mathematics necessary for his reasoning. He says he wanted to avoid a disproportionate and illegible text: a draft would have exceeded 250,000 pages. Lean, the language used for this release, is an open-source programming language.
Specialists will still have to examine the demonstration independently of its promoter. The proposal is also set to compete for the Millennium Prize, which includes the Riemann hypothesis among seven mathematical problems and provides a $1 million prize. No payment would be immediate: the settlement requires review for two years.
The evening letterFree · 7 p.m.
AI transforming business?
AI strategies, movements and tricks deciphered, every evening in 5 minutes. Free.
Included upon registration: our selection of the best AI guides & comparisons.
Choose your pace
Free · No spam · Unsubscribe in 1 click
Alpöge attributes the outcome to Claude
A number theory specialist and Harvard graduate, Alpöge made his claim public on his personal website. The hypothesis concerns prime numbers and a possible underlying order. The researcher says he collaborated with Claude during the last months and attributes the decisive role to the model in the final phase of his work.
The scope he attributes to his method goes beyond this single problem. According to him, it also makes it possible to demonstrate the generalized Riemann hypothesis. He hopes to then be able to deal with the extended Riemann hypothesis, without claiming to have already obtained this result.
Other mathematical work related to AI
Other results attributed to the models include a proof of Fermat’s Last Theorem by Anthropic, originally stated in 1637, and a Google Gemini discovery of a 59-year-old nuclear fusion conjecture. OpenAI, for its part, published a proposal on the Navier-Stokes equation. Some academics have described the corresponding series of publications as a “giant bombshell”.
Bryna Kra, a professor of mathematics at Northwestern University, recently expressed at MIT her astonishment at the rapidity of the changes. According to Kra, the last two months have transformed mathematics more than the last 100 years. This appreciation is accompanied by a feeling combining enthusiasm and fear.

