aipopdb

22 Nov 2026  ·  10,137 works · 1,192 news indexed

News / Anthropic mathematician claims Claude-assisted Riemann hypothesis proof, still unverified

Anthropic mathematician claims Claude-assisted Riemann hypothesis proof, still unverified

AIPopDB Newsroom · via scientificamerican.com · · 3 min read · Comment

Read the original ↗
Image for Anthropic mathematician claims Claude-assisted Riemann hypothesis proof, still unverified

vectorfusionart/Shutterstock · Image: Scientific American

Watch on YouTube ↗

Levent Alpöge, a mathematician who works at Anthropic, said on his personal website on 20 November 2026 that he believes he has found, with the help of Claude, a proof of the Riemann hypothesis. His statement, about four pages long, says the argument also covers the generalized version of the hypothesis. Anthropic had issued no announcement of its own when the first reports appeared, and the claim has not been confirmed by other mathematicians.

“If the proof is correct and understood, this completes a dictionary between different branches of math where the connections were only partial”

Hector Pasten, mathematician at the Pontifical Catholic University of Chile, to Scientific American · scientificamerican.com ↗

The hypothesis dates from 1859, when Bernhard Riemann proposed it, and it governs the seemingly random pattern of the prime numbers. Hundreds of published theorems assume it is true, and the Clay Mathematics Institute offers a $1 million Millennium Prize for a proof. Alpöge has released a version of the argument written in Lean, a language that checks each logical step mechanically, running to millions of lines of code, but there is no written paper yet. Reports put the claimed proof at more than 250,000 pages. Andrew Sutherland of MIT says he compiled the Lean file himself. Lean can still go wrong if the hypothesis is encoded incorrectly, so experts will need to check both the encoding and the mathematics.

Mathematicians quoted so far are cautious but interested. Sutherland called it a historic moment and a human triumph, while saying that understanding the details will take time. Hector Pasten, of the Pontifical Catholic University of Chile, gave a conditional welcome, on the premise that the proof is correct and understood. Carlo Pagano said that from a first look the ideas could open new routes in number theory. Alpöge himself said the progress amazed him and that he expects Claude may one day prove an even broader version, while acknowledging a sense of loss among mathematicians.

Futurism's lead illustration for its report on the claim

Futurism's lead illustration for its report on the claim

Image: Futurism

The claim follows a summer of AI results in mathematics. In July, Alpöge shared a Claude-generated result on the Jacobian conjecture, and in September OpenAI made a claim on the Navier-Stokes problem, after which NYU's Tristan Buckmaster implied that OpenAI had used his and Alpöge's work, setting off a dispute over credit. Reports differ on Alpöge's affiliation: two describe him as an Anthropic employee, and one as a Harvard mathematician.

About Levent Alpöge

Levent Alpöge is a mathematician who works at the AI company Anthropic; Scientific American describes him as a mathematician and Anthropic employee.

In July 2026 he posted a counterexample to the Jacobian conjecture that he had found with Claude: a polynomial map from three-dimensional space to itself whose Jacobian determinant is constantly -2, which disproves the conjecture for three dimensions and, trivially, for all higher ones. In September 2026 he was at the centre of a credit dispute over the Navier-Stokes problem, after NYU's Tristan Buckmaster implied that OpenAI had taken his and Alpöge's work.

For now it remains an unverified claim. One report adds that the proof has no immediate practical uses, such as breaking bank security, and the next step is for other experts to examine the Lean file and the mathematics behind it.

Comments