Anthropic Mathematician Levent Alpöge Claims Riemann Hypothesis Proof With AI

Anthropic mathematician Levent Alpöge said on November 20 that he and the company’s Claude AI model have produced a proof of the Riemann hypothesis, a 167-year-old problem that carries a $1 million Millennium Prize.

The claim arrived the day after Anthropic’s initial public offering. Alpöge acknowledged the timing was awkward. “The precise timing is horrible,” he said. “If I could have, I’d have finished this all much earlier.”

Mathpocalypse: Riemann Hypothesis "solved" by Claude AI and Anthropic mathematician

Rival proof appears hours later

Hours after Alpöge’s announcement, Stanford postdoctoral researcher Alessandro Masullo posted his own proof — one he said he had pursued for 12 years without AI. “In light of the current climate I am being advised to release the current version as it is,” Masullo wrote, noting it may contain errors because it was rushed.

“If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere.”

“Now it’s getting more complicated,” Caltech professor Sergei Gukov told Business Insider. Cornell professor Steven Strogatz called the situation “The plot thickens!” on X. Neither proof had been formally verified by Friday afternoon.

Proof released only in Lean code

Alpöge released his argument only in Lean, a formal proof language that mechanically checks each logical step. He said a conventional write-up would exceed 250,000 pages and deemed it unreadable. “I decided to only present Lean,” he wrote, explaining he considered a massive generated document inappropriate for publication.

Branding for Claude Code on removable posters at Anthropic PBC's Code with Claude developer conference in London
Photo: The Economist

Independent researchers have begun compiling the Lean code. Andrew Sutherland at MIT reported it compiled without issues using only the three standard axioms. Kevin Buzzard of Imperial College London said a “comparator run” is still needed to confirm the code does not redefine symbols or exploit a kernel bug. “Honestly the chance that anything is fishy here is approximately zero,” Buzzard wrote on a Lean community chat server.

Method targets generalized hypothesis

Alpöge claims the method resolves the Generalized Riemann Hypothesis, a broader conjecture extending the same prediction to a larger family of mathematical functions. He also believes the approach could reach the Extended Riemann Hypothesis.

In August, Anthropic said an unreleased research version of Claude had raised a longstanding lower bound for the proportion of zeta zeros on Riemann’s predicted line from 41.6% to 67.2%, while explicitly stating that Claude had not proved the hypothesis.

Verification remains years away

Terence Tao, a prominent mathematician, said verification is still some way off. “I imagine the digestion process will take some time, though in this case at least we have a human author to communicate with to assist with this task,” Tao said. Under Clay Mathematics Institute rules, a solution must be published in a peer-reviewed journal and withstand two years of scrutiny before the $1 million prize is considered.