An Anthropic mathematician claims to have proven the Riemann hypothesis with Claude. Verification, however, will take years Home News Mathematician Levent Alpöge, who works for Anthropic, claims to have found a proof of the Riemann hypothesis with the help of Claude The 1859 hypothesis is one of the seven Millennium Prize Problems, for the solution of which the Clay Mathematics Institute has offered a million-dollar reward The proof has not yet been independently verified. To be eligible for the prize, it must withstand at least two years of public scrutiny Sdílejte: Jakub Kárník Published: 22. 11. 2026 01:30 Advertisement Announcing a proof and having a recognized proof are two different things in mathematics. Levent Alpöge from Anthropic this week published a claim that he solved the Riemann hypothesis with Claude. It has resisted even the world’s best mathematicians for decades. Verified on trillions of solutions, yet unproven The hypothesis was formulated by Bernhard Riemann in 1859 and concerns the distribution of prime numbers. According to the Clay Mathematics Institute, it states that all “interesting” solutions to the equation ζ(s) = 0 lie on a single vertical line. This has been computationally verified for the first 10 trillion solutions. However, checking a finite number of cases is not proof for all. Alpöge, however, claims more than just a proof of the hypothesis itself. According to him, his method also solves the so-called generalized Riemann hypothesis and could eventually handle an extended version. On his website, he adds that he formalized the arguments with Claude without assumptions. He published the proof in the Lean language. How many prime numbers are there between one and a million? Gauss’s formula estimates their number to be around 78,628. In reality, there are 78,498, so the estimate was off by only 130. The Riemann hypothesis states how large such a deviation can be, even if, instead of a million, we choose a number so enormous that we can no longer count prime numbers one by one. If the proof were confirmed, mathematicians would gain a more reliable bound for the error of their estimate. At the same time, they could use results that currently hold only under the assumption that Riemann was correct, without further conditions. Prime numbers have specific uses, for example, in an Android phone. When you turn it on, it can use an RSA digital signature to verify that it is running the manufacturer’s system and that no one has altered it along the way. To create a signing key, the manufacturer needs two enormous prime numbers. How does a computer know that a randomly selected number with hundreds of digits is truly a prime number? If the generalized Riemann hypothesis, which Alpöge also claims to have proven, were confirmed, a specific procedure called the Miller test would gain a mathematical guarantee of how many checks are always sufficient for a correct answer. This is where the connection to computational technology lies. The benefit would be the certainty that the algorithm will not make a mistake even with numbers that no one has tested yet. Anthropic is silent, prize decision at least two years away Anthropic itself has not commented on the announcement, according to The Verge, and has not been involved in it. Yet, in August, the company stated that it was working on the hypothesis with Claude. At the same time, however, it claimed that it did not expect Claude’s techniques to lead to a proof. The timing is also interesting. According to his document, Alpöge wrote about the solution to company CEO Dario Amodei and apparently waited to publish until after Anthropic’s Nasdaq debut on Thursday. The Clay Institute has not yet publicly accepted the proof. According to its rules, the solution must be published, must withstand scrutiny for at least two years, and gain general acceptance within the mathematical community. Artificial intelligence contributes to mathematics, but no official recognition yet In the quarter-century since the prize was announced, only one Millennium Problem has been officially solved: the Poincaré conjecture. In September, OpenAI announced a solution to the Navier–Stokes equations. This also lacks formal recognition so far and is accompanied by accusations that the company used unpublished work by other mathematicians. Alpöge himself stated in July that he had solved Jacobi’s conjecture with Claude. Therefore, it is not a confirmed result, but a claim that now needs to be verified by other mathematicians. Formalization in Lean may facilitate their work, but the decisive word will belong to the community, not the author. Should a proof developed in collaboration with AI be judged the same as one by a human, or more strictly? Sources: tomshardware.com, theverge.com About the author Jakub Kárník Jakub is known for his endless curiosity and passion for the latest technologies. His love for mobile phones started with an iPhone 3G, but nowadays… More about the author Sdílejte: Anthropic Claude matematika Umělá inteligence You might be interested in The Gemini app targets weaker phones with Android Go. Google rebuilt it for them Jakub Kárník 21. 11. Google introduced Gemini 4 Neon. It will offer the performance of top-tier models, just faster and cheaper Adam Kurfürst 19. 11. OpenAI released GPT-6.2 Astra and Sol. Both models are more powerful, and Astra is cheaper Jakub Kárník 17. 11. AI at Hand #4: Don't know what Christmas gift to buy? Gemini will tailor one for you after five questions Adam Kurfürst 17. 11. Would you trust AI with your credit card? ChatGPT Wallet lets agents shop for you Adam Kurfürst 15. 11. A bug in ChatGPT released Dots agents into Europe. A technical error was behind it Adam Kurfürst 13. 11.