Did Anthropic’s Claude Just Help Crack Math’s 167-Year-Old Riemann Hypothesis?

On Friday, November 20, the mathematics and AI worlds collided in dramatic fashion. Levent Alpöge, a mathematician at Anthropic, announced that he had developed a proposed proof of the Riemann Hypothesis—widely regarded as the most famous unsolved problem in mathematics—with the help of Anthropic’s Claude AI model.

First proposed in 1859 by German mathematician Bernhard Riemann, the conjecture has stumped generations of human geniuses for 167 years and carries a $1 million Millennium Prize from the Clay Mathematics Institute. Even more striking, Alpöge said the method could extend to two broader versions: the Generalised and Extended Riemann Hypotheses.

What the Claim Actually Is—and Why Caution Is Essential

Alpöge did not just drop a PDF; he posted his Claude-assisted proof formalized in Lean, a programming language designed to mechanically verify mathematical arguments. Still, Lean verification only checks that the logic matches the way the problem was coded—it does not guarantee the formalization itself is free of subtle gaps, and Alpöge himself acknowledged the argument is so long that he has not yet fully digested every piece of it.

  • Why It Matters to Math: Prime numbers look random on the number line, yet Riemann predicted a hidden order governing where the non-trivial zeros of the Riemann zeta function sit. Proving it would reshape number theory and cryptography.
  • How Claude Contributed: After months of collaboration, Alpöge wrote that “Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line,” calling the moment “one dawn across science of something far greater.”

AI’s Growing—and Contested—Role in Frontier Math

The announcement lands amid a broader surge in AI-assisted mathematics. In September, OpenAI claimed an internal model produced a proposed solution to the Navier–Stokes problem, and Claude previously formalized Fermat’s Last Theorem in Lean. At the same time, some mathematicians remain deeply skeptical, warning that AI-generated proofs risk flooding journals with massive, opaque arguments that take human experts years to audit.

Whether Alpöge’s proof survives independent peer review or joins the long list of near-misses, one thing is clear: AI has moved from solving textbook exercises to knocking on the door of mathematics’ deepest mysteries.

Scroll to Top