Levent Alpöge Releases Lean Code for Claude-Assisted Riemann Proof Claim

News Summary
On Friday, November 20, 2026 (Eastern Time), mathematician Levent Alpöge, who works at Anthropic and earned a mathematics Ph.D. from Princeton University, posted a four-page statement on his personal website saying that he believes he and Claude have found a proof of the Riemann Hypothesis. He said the approach also establishes the Generalized Riemann Hypothesis for Dirichlet L-functions, and released a large Lean formalization rather than a conventional, readable research paper. The announcement is a claim, not yet an accepted resolution: independent mathematical scrutiny of the argument and its formal statements remains essential.
Why the Riemann Hypothesis matters
First proposed by Bernhard Riemann in 1859, the hypothesis concerns the nontrivial zeros of the zeta function. It predicts that each lies on a particular vertical line in the complex plane, whose real coordinate is one-half. That statement is deeply connected to the distribution of prime numbers: proving it would give mathematicians much sharper control over how prime counts differ from their expected estimates.
The Generalized Riemann Hypothesis extends the idea to Dirichlet L-functions, which help describe primes in arithmetic progressions. A proof of that broader statement would have consequences across number theory. Both claims are exceptionally ambitious, and the classical hypothesis remains listed by the Clay Mathematics Institute as an unsolved Millennium Prize Problem while the new claim is examined.
What Alpöge announced
Alpöge's note says he worked with Claude over recent months. He describes guiding the effort toward an analogy with methods used to study zeta functions over finite fields, while Claude proposed key steps in an approach involving chains of theta lifts. The note presents the Generalized Riemann Hypothesis as within the claimed result and says the Extended Riemann Hypothesis might be a future direction.
The author did not present a standard proof manuscript. He wrote that a fully detailed prose account would be extraordinarily long and difficult to read, and said he intends to understand the argument more deeply, simplify it, and write it up. The public release instead consists of a short explanatory note and a multi-part archive of Lean code.
What Lean can establish
Lean is a proof assistant: a programming language and small trusted checking system used to encode mathematical definitions and proofs. If a Lean development builds successfully, the kernel checks that the formal steps follow from the stated assumptions and definitions. This can provide a powerful safeguard against gaps in a formal derivation.
That check does not, by itself, establish that the formal statement captures exactly what a researcher intended, that imported definitions and dependencies are appropriate, or that people understand why the result is true. Reviewers must inspect the theorem statements, assumptions, dependency chain, and proof development, and compare them with the standard mathematical formulations. A successful build is important evidence about formal validity, but it is not a substitute for expert examination of the mathematics.
The next stage of review
The release invites technically capable readers to compile the archive and inspect its main declarations. Independent checks should verify that the Lean theorem for the classical hypothesis and the statements for Dirichlet L-functions match the intended versions, and that no unexamined assumptions enter through libraries or dependencies. A readable account would also help number theorists evaluate the mathematical strategy and identify simplifications or gaps.
As of Saturday, November 21, 2026 (Eastern Time), the announcement should therefore be treated as an extraordinary research claim under review, rather than a confirmed solution. If the formalization and its interpretation withstand independent scrutiny, the work could become a landmark in both number theory and computer-assisted mathematics. If not, the checking process may still yield useful formal techniques, intermediate results, or lessons about how AI tools can assist research.
A case study in AI-assisted mathematics
The announcement highlights a developing partnership between mathematical expertise and language models. Alpöge brought years of preparation and made choices about the mathematical direction; Claude is credited with contributing ideas and helping formalize the argument. That combination offers a useful educational example of how AI can participate in research while leaving humans responsible for defining questions, checking claims, and making results understandable.
For students, the episode also shows why a mathematical breakthrough involves more than producing code. It requires precise definitions, verifiable reasoning, careful communication, and a community able to test the result. Lean can make a proof's formal structure inspectable, while a clear exposition can show researchers how its ideas fit together. Both forms of scrutiny matter when a claim reaches beyond a single specialist.
This article was compiled by the AIBARS editorial team with AI assistance. AI can make mistakes, so please check the original source for anything important. Spotted an error? Email [email protected].