
An Anthropic mathematician says he and Claude may have proved the Riemann Hypothesis, one of mathematics’ most famous unsolved problems. The claim is attracting international attention, but it is not yet an accepted mathematical result. Independent experts are checking whether the computer code proves the intended statement and whether the argument withstands scrutiny.
Levent Alpoge, a member of Anthropic’s technical staff, published a four-page note on 20 November saying he believes he has discovered a proof while working with Claude. He said the approach also extends to the Generalised Riemann Hypothesis, a broader claim. He released an extensive formalisation in Lean, software used to check mathematical proofs, rather than a conventional human-readable paper explaining the full argument.
According to Quanta Magazine’s reporting, several independent researchers have successfully compiled the Lean code, which runs to about 17 million lines. That is an important first step, but it is not the same as the mathematical community accepting the claimed proof. Further checks must establish that the formalised theorem is the Riemann Hypothesis as ordinarily understood and that no definition or assumption has changed the question being answered.
Quanta reported that comparator checks were still under way when it updated its account. Some mathematicians also want a clear conceptual explanation of why the proposed approach works. Alpoge acknowledged in his note that he has not yet mastered every part of the very large argument and intends to write it up properly. Anthropic had not made a formal company announcement at the time of Quanta’s report.
The hypothesis, formulated in the nineteenth century, concerns the zeros of the Riemann zeta function and has deep consequences for understanding how prime numbers are distributed. It is a Millennium Prize Problem, a designation that helps explain why even a plausible claimed solution draws immediate scrutiny. The broader Generalised Riemann Hypothesis would reach beyond the original statement, making the scope of Alpoge’s announcement especially striking if it is validated.
This story is also different from earlier Claude-assisted work on a weaker result. That research did not establish the full Riemann Hypothesis. The new claim must be judged on its own published materials and independent review, not on the reputation of the AI model or the excitement around it.
If the formal proof survives checking, the collaboration would be a remarkable example of an expert using AI in advanced research. It would not mean a chatbot solved a century-old problem unaided. Alpoge describes a sustained collaboration in which his mathematical judgement directed the work, while Claude contributed technical ideas and formalisation.
The immediate news is the publication of a claimed proof and the race to verify it. Associated Press reported the claim early on Saturday. Until independent mathematicians complete their examination and the result earns broader acceptance, headlines should not say that the Riemann Hypothesis has been solved.







