Levent Alpöge, a 34-year-old Turkish-American mathematician at Anthropic, announced on Friday that he had proved the Riemann hypothesis using the company's Claude AI [1]. The Lean and Mathlib communities have since confirmed the proof is correct [2].
Alpöge published the proof on his website on Nov. 20 as 17 million lines of code in Lean, a programming language that checks mathematical proofs for errors [2]. Dozens of people ran the code independently and got the same result, according to a report by MIT mathematician Andrew Sutherland [3]. The Lean Focused Research Organization and the Mathlib Initiative issued a joint statement on Saturday saying, "We have full confidence that Alpöge's proof is correct" [4].
The Riemann hypothesis, formulated by Bernhard Riemann in 1859, concerns the distribution of prime numbers—numbers divisible only by 1 and themselves [1]. It predicts that the difference between the actual and estimated count of primes below a given number is no larger than the square root of that number. The hypothesis is one of the seven Millennium Problems selected by the Clay Mathematics Institute in 2000, each carrying a $1 million prize [5].
Alpöge, who earned his PhD at Princeton under Fields Medalist Manjul Bhargava, said he worked with a non-public version of Claude for about six weeks to develop the proof [6]. In a statement on his website, he described the AI-written proof as "incredibly long," running to more than 250,000 pages in human-readable form [5]. "Claude flew, danced and shone just before the finish line," he wrote [1].
The proof extends beyond the classical Riemann hypothesis to the generalized Riemann hypothesis for Dirichlet L-functions, which encode the distribution of primes in arithmetic sequences [3]. The verification confirmed that the code matches the standard formulation of the hypothesis and contains no additional assumptions beyond the standard logical foundations [3].
The announcement came one day after Anthropic's IPO on Nov. 19, which valued the AI company at $1.5 trillion. Alpöge denied the timing was related, writing that he had wanted to publish several days earlier but did not receive permission from Anthropic's CEO [1]. Oxford mathematician James Maynard told NRC on Friday night that he estimated the chance the proof is correct at "more than 99 percent" [1].
Even Alpöge does not fully understand the proof. "I intend to do my best to understand this proof and its certain simplifications and to make them understandable to others," he wrote [5]. The question of why the Riemann hypothesis is true remains unanswered, according to NRC [1].
Forecasters on Metaculus had put the expected date at May 2035. Mathematician Terence Tao said last month he expected the problem to fall to "a combination of humans and AI around 2050 or so" [2]. Hours after Alpöge's announcement, independent mathematician Alessandro Masullo published a 940-page sketch titled "The Music of the Primes." He claimed an incomplete proof he had worked on for nearly 12 years without using AI [7].

