Anthropic mathematician claims proof of the Riemann hypothesis with Claude
Anthropic technical staff member Levent Alpoge announced on Friday that he and Claude discovered a proof of the generalised Riemann hypothesis, extending Bernhard Riemann’s 1859 conjecture to a broader family of functions. Posted on Alpoge’s website with a four-page explanation, the claimed proof would settle a 167-year-old question about prime number distribution encoded in the Riemann zeta function, one of the Clay Mathematics Institute’s $1m Millennium Prize Problems.
Rather than a conventional paper, Alpoge published 17 million lines of code in Lean, a proof-assistant programming language that checks reasoning step by step—potentially tens of thousands of pages of mathematics. He wrote that Claude "far outpaced me near the finish line."
Independent experts are trying to verify the argument, though no human has yet read it in full. Imperial College London professor Kevin Buzzard said researchers worldwide are translating the Lean code into English to see how much is genuinely new, adding that experience suggests humans should understand what is happening in the code within a week or two.
Queen Mary University of London professor Abhishek Saha noted that hundreds of mathematical results already assume the hypothesis holds. Buzzard added that the generalised hypothesis also underpins fast algorithms used in internet security and cryptography, giving it practical significance beyond OpenAI’s September claim of solving the Navier-Stokes problem.
Continue in the app — vote & join in ➔