Anthropic's Claude advances the Riemann hypothesis: zeta bound raised from 41.6% to 67.2%
(Published on Anthropic's research site on 2026-10-05; reported via Ground News aggregation on 10-07)
Key points
- What was done: An unreleased research version of Claude from Anthropic raised a zeta bound related to the Riemann hypothesis from 41.6% to 67.2%. Mathematicians had chipped away at this bound bit by bit for decades; the AI lifted it a long way in one step.
- Credibility: Anthropic does not claim to have solved the Riemann hypothesis, and explicitly limits the result to an improvement of a numerical bound. The result was mechanically checked in the Lean formal proof language and published only after review by in-house mathematicians.
- Background: Less than two weeks ago, Anthropic completed the first full computer formalization of Fermat's Last Theorem; a few days ago, OpenAI released 722 papers at once, covering 377 open math problems.
- Why it matters: Mathematics is becoming the new arena for frontier AI labs, replacing coding leaderboards — the contest is over whose model can do real work that human mathematicians recognize.
Source: read the original. This text is a summary for reference only.
