Anthropic researcher Levent Alpöge claims Claude AI solved Riemann hypothesis
An artificial intelligence model and a researcher have discovered a proof for the century-old Riemann hypothesis, releasing a computer-formalized Lean version.
An artificial intelligence model assisted in cracking a century-old mathematical mystery, raising profound questions about the future of human research as autonomous systems accelerate discovery.
Levent Alpöge, a number theorist and researcher at Anthropic, announced on November 20, 2026, that he and the artificial intelligence system Claude had discovered a proof for the Riemann hypothesis. First formulated in 1859 by German mathematician Bernhard Riemann, the conjecture explores the distribution of prime numbers and governs the zeros of the zeta function. For generations, the problem has stood as the most famous unsolved challenge in number theory, commanding a $1 million bounty from the Clay Mathematics Institute as one of its seven designated Millennium Prize Problems.
Media additions
Alpöge published his findings through a personal website, releasing a computer-formalized version of the argument written in Lean, an open-source programming language and proof assistant that checks logical deductions through a kernel. According to Alpöge, an uncompiled human-readable draft of the work exceeded 250,000 pages, prompting him to bypass a traditional manuscript in favor of machine code to avoid publishing an unmanageable text. He stated on his website:
"I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis. In fact, the method easily extends much further, resolving the so-called Generalized Riemann Hypothesis, and I expect, with more time, the Extended Riemann Hypothesis too."
Levent Alpöge, Anthropic researcher, via Tech. Yahoo
The announcement arrived on the heels of Anthropic’s public stock market debut, prompting the researcher to acknowledge potential cynicism regarding the timing of the release. Anthropic itself had previously stated in August that it did not expect its work with Claude on the Riemann problem to produce a complete proof, and the company did not formally endorse the announcement, maintaining distance from the researcher's independent release.
Reactions across the mathematical community were swift and mixed. Massachusetts Institute of Technology mathematician Andrew Sutherland independently compiled the shared Lean modules, reporting that all 175,187 code components successfully passed kernel verification and relied exclusively on permitted mathematical axioms. Sutherland described the development as an historic turning point for science, characterizing it as a human triumph built on millennia of preceding academic foundations. Other academics emphasized the sweeping implications of resolving the Generalized Riemann Hypothesis. Hector Pasten, a professor at the Pontifical Catholic University of Chile, likened the verified connections to a Rosetta Stone bridging disparate mathematical domains, while Concordia University professor and Google DeepMind visiting researcher Carlo Pagano noted that the underlying arguments could unlock entirely new avenues of thinking in number theory.
| Milestone / Claim | Involved Parties | Verification Status |
|---|---|---|
| Riemann Hypothesis Proof | Levent Alpöge and Claude | Pending community review; Lean code compiled by independent auditors |
| Generalized Riemann Hypothesis | Levent Alpöge and Claude | Claimed within formal release; requires expert evaluation |
| Navier-Stokes Solution | OpenAI / Tristan Buckmaster and Levent Alpöge | Credited by peers but lacks formal institute recognition |
| Jacobian Conjecture Counterexample | Levent Alpöge and Claude | Distributed via personal release |
The breakthrough follows a turbulent year for automated reasoning in academic circles. Earlier in the year, Alpöge published a counterexample to the Jacobian conjecture using Claude, and collaborated with New York University mathematician Tristan Buckmaster on the Navier-Stokes equations. That fluid dynamics work later became embroiled in controversy when OpenAI announced its own solution following allegations regarding shared research methods. Unlike cryptography or global financial networks, experts emphasize that a validated proof of the Riemann hypothesis carries no immediate technological application for banking systems or internet infrastructure, remaining strictly an achievement of pure mathematics.
Substantial hurdles remain before the claim achieves formal acceptance. Under the governing regulations of the Clay Mathematics Institute, any proposed solution must appear in a recognized publication, withstand at least two years of rigorous public scrutiny, and secure a consensus of approval from the global mathematical community before a prize can be awarded. As of late November, the institute's official registry continued to list the problem as completely unsolved.
What to Watch Next
- Independent research teams finishing end-to-end sandbox evaluations of the published Lean modules.
- Formal commentary and peer-reviewed critiques from international number theory specialists.
- Future updates from the Clay Mathematics Institute regarding the two-year observation window required for prize consideration.