Home » Anthropic’s Levent Alpöge Claims AI-Assisted Proof of Riemann Hypothesis

Anthropic’s Levent Alpöge Claims AI-Assisted Proof of Riemann Hypothesis

By

–

Anthropic's Levent Alpöge Claims AI-Assisted Proof of Riemann Hypothesis

Anthropic mathematician Levent Alpöge announced on November 20 that he believes he and the company’s AI model, Claude, have discovered a proof of the Riemann Hypothesis, a problem that has resisted solution since 1859. Instead of a traditional paper, Alpöge released 14 parts of Lean formalization code totaling 17.37 million lines, a volume driven by an AI-generated draft that exceeded 250,000 pages.

The North Star of Number Theory

The Riemann Hypothesis concerns the distribution of prime numbers — those integers divisible only by one and themselves, like 2, 3, 5, and 7. In 1859, Bernhard Riemann connected the primes to the behavior of the zeta function, a complex mathematical object. The hypothesis asserts that all the “nontrivial zeros” of this function lie on a single vertical line in the complex plane, where the real part is exactly one-half. This is not abstract curiosity; the Clay Mathematics Institute explains that while the prime number theorem describes the average distribution of primes, the Riemann hypothesis governs the deviations from that average. It has been checked for the first 10 trillion zeros, but a proof for all of them remains the open problem.

Why It Is the ‘Most Important’ Unsolved Problem

The hypothesis acts as a structural keystone for vast swathes of mathematics. It is the most famous of the seven Millennium Prize Problems, carrying a US$1 million bounty from the Clay Mathematics Institute for a qualifying solution. Progress has been so rare that top experts historically spent little time directly attacking it, because no obvious path to a proof existed. As Scientific American reported, it serves as the “North Star for the entire field of number theory.” A confirmed proof would instantly validate a “deluge of truths” contingent on the conjecture, completing a dictionary between disconnected branches of mathematics where connections were previously only partial.

The AI-Assisted Method

Alpöge describes a six-week collaboration where he guided the initial mathematical direction before Claude accelerated rapidly near the finish line, “flew away” from him, and generated an argument he admits he does not yet fully understand. The proof leans on deep machinery: mathematicians like Chris Birkbeck at the University of East Anglia note it uses Galois representations and automorphy-lifting arguments involving base change, Jacquet–Shalika, and Deligne–Ribet theorems. Alpöge claims the method also resolves the Generalized Riemann Hypothesis and expects it could extend to the Extended Riemann Hypothesis with more time.

Verification and the Road Ahead

The Lean code has already compiled successfully for several independent verifiers, including MIT mathematician Andrew Sutherland. However, compilation checks only the logical syntax, not the mathematical meaning of the definitions and axioms used. The Clay Mathematics Institute’s problem page still listed the hypothesis as unsolved as of Friday. Institute rules require publication and acceptance by the mathematics community, followed by a two-year scrutiny period, before any prize is awarded. Alpöge intends to understand the argument more fully, explain it in talks, and prepare a human-readable presentation. He described the discovery as “one dawn across science of something far greater.”

Sources

Photo: Emile Perron / Unsplash (Unsplash License)

Get breaking stories first on our Telegram channel.

More stories by

Most read

Crime & Fraud

People in the News

News

World News