--:--:-- ET
New York 62°F Partly cloudy Los Angeles 78°F Sunny Chicago 55°F Windy
Subscribe
U.S. Edition Est. 2026 Nov. 21, 2026

US News America

Independent reporting for a changing America

Breaking
Technology

Anthropic Mathematician Says AI Proved the Riemann Hypothesis, and Stanford Posts a Rival Human Proof Hours Later

Levent Alpoge said he and Anthropic's Claude produced a proof of math's most famous open problem, released as computer code. Hours later, Stanford posted a competing proof by a young mathematician. Neither has been verified yet.

A chalkboard covered in handwritten mathematical equations and prime numbers, with a hand holding chalk at the edge

A mathematician at Anthropic said on Friday that he and the company’s artificial intelligence model Claude have produced a proof of the Riemann hypothesis, the most famous open problem in mathematics. Hours later, a second claim arrived from Stanford, where a young mathematician said he had written a proof of the same problem without any help from AI.

Levent Alpöge, a member of Anthropic’s technical staff, posted his announcement on his personal website. “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” he wrote. The Riemann hypothesis, first proposed in 1859 by the mathematician Bernhard Riemann, concerns the pattern in the distribution of prime numbers. It carries a $1-million prize from the Clay Mathematics Institute, which still listed the problem as unsolved hours after Alpöge’s announcement.

Alpöge said his proof also settles the Generalized Riemann Hypothesis, a broader version of the problem. He released the argument as computer code in Lean, a programming language that checks each step of a proof for correctness. The Associated Press reported the proof runs to 17 million lines of Lean code, and Alpöge said a written version would fill more than 250,000 pages.

A computer monitor showing lines of programming code on a desk at night, with a keyboard and a cup of coffee

A second claim, written by a person

Hours after Alpöge’s post, Ravi Vakil, president of the American Mathematical Society and a mathematician at Stanford University, posted a competing draft proof by Alessandro Masullo, a young mathematician at Stanford. Scientific American reported that Masullo’s proof had been in the hands of Stanford mathematician Brian Conrad, an expert on the subject, who had been giving Masullo feedback for months. Conrad confirmed the details of Vakil’s post to Scientific American and said more work was needed, adding that the paper was undergoing major revision.

According to Scientific American, Masullo had planned to spend a year at Stanford refining the draft before going public, but Anthropic’s announcement forced the researchers to release the current version at once.

A stack of mathematics textbooks and printed papers on a wooden desk beside an open laptop

Neither proof has been verified. Mathematicians around the world began checking Alpöge’s code within hours of its release, and the size of the archive means the checking will take time. Andrew Sutherland, a computational number theorist at the Massachusetts Institute of Technology who independently compiled Alpöge’s result in Lean, called it “an historic moment” and “a human triumph,” while adding that much work remains to understand the details of the proof.

If either proof holds up, the effect on mathematics would be large. Many results in number theory depend on the Riemann hypothesis being true, and a proof would settle them at once. “If the proof is correct and understood, this completes a dictionary between different branches of math where the connections were only partial,” Hector Pasten, a mathematician at the Pontifical Catholic University of Chile, told Scientific American, “something like a Rosetta Stone.”

A spiral of glowing white dots on a dark blue background, an abstract image of prime numbers

The announcement came a day after Anthropic’s record-breaking debut on the stock market. Alpöge said the final pieces of the work came together late on Sunday night, November 15, after months of work with Claude in which he set the direction and the model carried out the detailed steps. He had treated the problem as a lifelong goal since college. In October, results from OpenAI had already made progress on the prime numbers, and in July Alpöge posted a Claude-generated solution to the Jacobian conjecture, another long-standing problem.

Reporting based on coverage by Scientific American, the Associated Press and The Wall Street Journal.