Intelligence Report

Anthropic mathematician claims Claude-assisted Riemann proof as Stanford rushes out rival draft

A Lean-verified AI proof of the Generalized Riemann Hypothesis and an all-human Stanford draft surfaced within hours, and mathematicians are now testing both.

INSIGHTEX Editorial Desk

2 min read

Anthropic mathematician claims Claude-assisted Riemann proof as Stanford rushes out rival draftIntelligence Report

What happened

Levent Alpöge, a mathematician at Anthropic, said on Friday that working with the company's Claude model he had found a proof of the Riemann hypothesis, the prime-number problem that has stood open since 1859, The Decoder reported. He put a paper and code on his website and said the approach also resolves the Generalized Riemann Hypothesis, with the argument written in the Lean language so a computer can check it.

Within hours, Stanford professor and American Mathematical Society president Ravi Vakil used his blog to share a separate draft proof of the Generalized Riemann Hypothesis by mathematician Alessandro Masullo, which Vakil described as made without AI. Number theorist Brian Conrad had privately reviewed it for months. The plan had been a year of refinement, but the team published early after Anthropic's news. About eight hours later, Vakil said people had flagged problems, found with help from several AI models, and that the community must settle every issue before acceptance.

Alpöge said a printed version of Claude's argument would exceed 250,000 pages because the model writes out even elementary steps, so he released it only in Lean, split across 14 data packages. He acknowledged that Claude outpaced him in some subfields and said he intends to produce a human-readable version over the coming months. The claim arrived a day after Anthropic's IPO on Thursday; Alpöge called the timing unfortunate and rejected the idea that it was a publicity move.

Context

Claude has already been credited with several open math results in 2026, including a solution to the Jacobian conjecture that Alpöge published in July. In September, NYU mathematician Tristan Buckmaster suggested OpenAI had drawn on prior work by him and Alpöge on the Navier-Stokes equations, and in October OpenAI released prime-number results. The Riemann hypothesis has carried a $1 million Clay Mathematics Institute prize since 2000.

SourceThe Decoder

The briefing

The week in tech, in one email.

The stories worth your time, every week. Unsubscribe anytime.