The proof of the Riemann Hypothesis completed by Anthropic mathematician Levent Alpöge in collaboration with AI model Claude was certified by the Lean and Mathlib communities on November 21. According to the Lean community blogjoint statement, the two communities "fully believe that Alpöge's Riemann Hypothesis is proved correct" and formally certify that the propositions of this proof and the proof itself are correct.
The statement was jointly issued by the Lean and Mathlib communities and endorsed by the Lean FRO, which maintains the Lean programming language, and the Mathlib Initiative. Lean is a programming language that allows computers to check mathematical proofs line by line, and Mathlib is its main mathematical library. Lean’s official account is also on Xforward statementAnd congratulations to Alpöge.
The Riemann Hypothesis was proposed by the German mathematician Riemann in 1859. It is related to how prime numbers are distributed and is one of the seven millennium problems of the Clay Mathematics Institute. When Alpöge announced the proof on November 20, it was not given a paper, but a set of Lean code with a total of 17 million lines. Lian News reported at the timecontent of his statement, there was no third-party verification at that time.
Completed the verification in one day: dozens of people recompiled, and multiple sets of checkers accepted it
The statement said that the Lean and Mathlib communities mobilized over the past day to verify this proof, which was proposed on the Xena Project Blog of Imperial College London on November 12.Four inspection standardsAll comply. These four items are designed for AI-generated Lean code to prevent the code from changing questions or exploiting system vulnerabilities:
- The proposition is correct: the target of the proof is the existing definition of Riemann Hypothesis in Mathlib, which has been carefully reviewed by Lean and number theory experts.
- Able to compile: Dozens of people have reported successfully compiling the entire set of code and producing identical certificate export files.
- The proof is the original question: the community wrote another file that only quoted Mathlib, wrote the Riemann Hypothesis proposition separately, and then compared it with two independent tools, comparator and con-leche, and confirmed that the theorem proved by Alpöge is exactly the same as Mathlib's Riemann Hypothesis.
- The proof is valid: The exported proof was given to the Lean official core, con-leche, con-ron, and multiple sets of checkers written by third parties in different programming languages to rerun, and all passed.
The statement specifically mentioned that the two sets of checkers, con-leche and con-ron, come with formal proofs. Compared with set theory, they cannot accept wrong proofs, thus responding to concerns that "Lean's logic itself may have loopholes." All checks are run on export files that will not be executed to avoid proving that malicious programs are hidden in the code.
MIT verification page: All 175,000 modules passed, including the Generalized Riemann Hypothesis
The detailed description attached to the statement was compiled by MIT mathematician Andrew Sutherland.Verify profile page, he noted that the check was run using Claude Code. The page shows that the entire set of code, a total of 175,187 modules, was compiled successfully with zero errors; the independent core nanoda written in Rust took 27 minutes and 42 seconds to accept the entire proof on 80 execution threads, while con-leche accepted 1,468,418 declarations.
In addition to the Riemann Hypothesis, this proof also covers the generalized Riemann Hypothesis of all Dirichlet L functions, which is true for any modulus and any characteristics. The page also states the scope: Dedekind ζ functions, Heckian L functions and general automorphic L functions are not included in this examination.
Lean certification does not mean getting a million bonus
Lean certification deals with "a computer checking whether this certificate is correct", and no one can read it yet. Alpöge previously said that he and Claude tried to produce a human-readable version, and the result was more than 250,000 pages. Therefore, they only disclosed the Lean code first, and then compiled it into a paper that the mathematics community can understand. The Lean community also wrote at the end of the statement that it looks forward to the mathematical community's help in explaining "the new ideas in this proof."
The Clay Mathematics Institute’s Millennium Puzzle Prize is $1 million, but according to itsAward rules, the solution must be published in a qualified publishing channel, be published for at least two years, and be generally accepted by the global mathematics community before the institute will begin review. The Xena Project article also reminds that just because the Lean code passes the test, it does not mean that the text version written separately will prove to be correct.
As of press time, Anthropic has not issued an official statement on this certification.
this article Claude assisted in proving the Riemann Hypothesis, which was certified correct by the Lean mathematics community. first appeared in Chain News ABMedia。




