The race to solve the Riemann Hypothesis took an unexpected turn on November 20: Anthropic mathematician Levent Alpöge said he had completed a proof with the help of AI, and hours later Stanford postdoctoral researcher Alessandro Masullo published an independent solution. Neither proof has been formally verified. Mathematicians are examining both, in a test of what AI can—and cannot—contribute to mathematical research.
Masullo releases an unfinished paper early
Masullo said he had spent 12 years working on the Riemann Hypothesis and had expected to publish his results in about a year. After Alpöge announced his AI-assisted work, Masullo decided to release the current version sooner. He noted in the paper posted online that the rushed publication could mean it still contains errors.
Masullo’s proof is considerably shorter than Alpöge’s. That difference gives researchers another reason to check the arguments closely, but length alone does not determine whether a proof is correct. Masullo also used the release to underscore the continuing role of human mathematicians, saying the work was intended to convey that “we are still here” and that human mathematicians will not disappear.
Neither proof has been verified
As of the afternoon of November 20, neither Alpöge’s nor Masullo’s argument had been formally validated. Mathematicians are reviewing the work by hand, with AI tools also assisting in the checks. Cornell mathematics professor Steven Strogatz commented on X that “the plot thickens.” Caltech mathematics professor Sergei Gukov likewise said the situation had become more complicated.
The key question is not who was first to claim a solution, but whether each argument can withstand detailed scrutiny. Publication is not the same as acceptance by the mathematical community; until verification is complete, it remains unclear whether the Riemann Hypothesis has been proved.
Why the Riemann Hypothesis matters
The Riemann Hypothesis describes a predicted relationship in the distribution of prime numbers. It is one of the Millennium Prize Problems, and a successful proof carries a $1 million prize. OpenAI has previously claimed to solve another problem on the same prize list, the Navier–Stokes equation problem.
Gukov has called the Riemann Hypothesis one of mathematics’ most consequential problems, while Strogatz has described its difficulty by comparing it to “Moby-Dick.” The problem’s standing, combined with recent claims by AI companies about solving long-standing mathematical challenges, has drawn broad attention to Alpöge’s announcement. Researchers must now determine what role AI played in developing, structuring or checking a complex proof.
The debate turns to proof quality and human contribution
Masullo’s independent work adds a second path to a debate that initially centered on AI-assisted research. He said he had pursued the problem for years and brought forward his paper after the AI-assisted result was announced, while cautioning that the current version may contain errors. Alpöge has said he worked with Claude for several weeks to develop his solution.
Both texts remain under review. For now, there is no basis to determine which, if either, is correct, and neither unverified claim should be treated as a resolution of the problem. The outcome will depend on whether mathematicians confirm the arguments—not on whether their authors used AI or which paper appeared first.