Sign in

AI may have solved the most important problem in mathematics

Humans spent 167 years trying to prove the Riemann hypothesis

Published on: Nov 21, 2026, 16:30:40 IST
The Economist
As Prefered on
Share via
  • facebook
  • twitter
  • linkedin
Copy link
  • copy link

ON NOVEMBER 19TH Anthropic, an American AI company, held the largest IPO in history. Though financial markets have been troubled since, investors might take comfort in the latest display of the technology’s power. On November 20th Levent Alpöge, a mathematician at Anthropic, said that he had made use of Claude, the company’s AI model, to prove the Riemann hypothesis, widely considered to be the most important unsolved problem in all of mathematics. “If it fully checks out, it is clearly a historic milestone,” says James Maynard, a mathematician at the University of Oxford.

On November 20th Levent Alpöge, a mathematician at Anthropic, said that he had made use of Claude, the company’s AI model, to prove the Riemann hypothesis (Unsplash)
On November 20th Levent Alpöge, a mathematician at Anthropic, said that he had made use of Claude, the company’s AI model, to prove the Riemann hypothesis (Unsplash)

The hypothesis gets its name from Bernhard Riemann, a German mathematician. In 1859 Riemann introduced a new mathematical entity, known as the Riemann zeta function, which he believed was intimately connected to the distribution of prime numbers (those which are only divisible by one and themselves). If he was right, then the seeming chaos of the prime numbers betrayed a secret organising principle. Mathematicians have spent the 167 years since trying to prove the hypothesis.

As well as being central to the understanding of prime numbers, the Riemann hypothesis offers a bridge between many varied and seemingly disconnected mathematical subfields. Scads of other mathematical proofs and hypotheses have been proposed, which work only if one assumes that the Riemann hypothesis is true.

It is too early to definitively claim that the dominoes have fallen. Dr Alpöge’s proof, posted to his website in a file labelled “remark.pdf”, has yet to be checked by his peers. What lends weight to his claims, however, is that he has played a role in some of the high-profile AI-enabled maths breakthroughs that have flooded the field in recent months. Hundreds of long-standing open-problems have been solved since September, leaving mathematicians with thousands of pages of AI-generated proofs to pick through.

The announcement was unconventional in other ways, too. “Claude far outpaced me near the finish line,” he wrote. “No, Claude flew, and danced, and sparkled near that finish line.” Such flowery language was, however, accompanied by 17m lines of code in Lean, a programming language that checks the logical consistency of proofs. If this one holds up, Dr Alpöge and Claude will have solved more than the Riemann hypothesis. He also claims they have cracked its generalised form, which extends the problem to a wider family of mathematical functions.

Mathematicians will spend the weekend poring over the proof. Beyond the question of its validity, many will want to know how the AI reached its conclusion. The result does not come completely out of the blue. On October 6th OpenAI, another AI lab, announced that one of its models had solved the quasi-Riemann hypothesis, an important stepping stone towards the full Riemann hypothesis. But Jonathan Keating, another mathematician at the University of Oxford, describes the result posted this week as a “very big leap forward”. “Quasi-Riemann was itself a colossal shock and I never thought I’d see that proved in my lifetime,” he says. “But this is several orders of magnitude more shocking.”

Stay updated with the latest Technology News, gadget launches, app updates, artificial intelligence and digital trends. Find reviews, comparisons and useful insights from the world of tech.