The communities behind the Lean proof assistant and the Mathlib mathematics library say they have certified Levent Alpöge’s proof of the Riemann Hypothesis. Their checks covered both the theorem’s statement and the proof itself, rather than stopping at “code compiled, everybody go home.” Source
“We hereby certify that both the statement and the proof of the Riemann Hypothesis are correct,” their joint statement says. Source
According to the statement, Alpöge released the proof on November 20, 2026, as a Lean library containing 17 million lines of code. Source
Scientific American reported on November 20 that Alpöge, an Anthropic mathematician, claimed to have used the company’s Claude AI model to solve the problem. First posed by Bernhard Riemann in 1859, the hypothesis concerns a hidden pattern in the distribution of prime numbers. Source
More than a successful build
Lean lets researchers write mathematical proofs in a form software can check. Getting the code to compile, or build without errors, doesn’t by itself establish that it proves the intended theorem. Source
Scientific American’s earlier report described independent researchers successfully compiling Alpöge’s work, while theorem comparisons and checks by independent proof-checking programs were still pending. The community statement reports that those checks have passed. Source Source
- Repeatable compilation. Dozens of people reported compiling the codebase successfully and producing the same proof file for checking, according to the statement. Source
- The right theorem. Experts vetted the Riemann Hypothesis statement in Mathlib. Two independent comparison tools,
comparator and con-leche –compare, then confirmed that Alpöge’s theorem exactly matches that statement, the communities say. Source
- Independent proof checks. The statement says the proof passed checks by Lean’s core proof checker (its kernel),
con-leche, con-ron and third-party checkers written in different programming languages. These checks used a static copy of the proof, out of reach of hypothetical malicious code in the source files. Source
Claude’s role and what comes next
Neither the community statement nor Scientific American’s report details the division of work between Alpöge and Claude or identifies the model version used. The proof’s certification doesn’t establish that a chatbot solved the problem alone. Source Source
The communities also look forward to mathematical work explaining the proof’s new ideas. Seventeen million lines are not exactly a breezy explainer. Source
Comments