Anthropic pointed an unreleased Claude at the Riemann hypothesis. It did not solve it — it nearly doubled the best known bound.
On 10 August Anthropic published what an unreleased research version of Claude did when asked to attack the Riemann hypothesis. It did not prove it. What it did was raise the lower bound on the fraction of the zeta function's non-trivial zeros known to satisfy the hypothesis from 41.6% to 67.2% — by Anthropic's account the largest single improvement ever made to that number, which had crept upward over decades of human work. The run took two Claude Code sessions, roughly 60 subagents, about 31 million output tokens and some 2,400 shell commands. The mathematics combined recent results from Baluyot, Goldston, Suriajaya and Turnage-Butterbaugh with Bombieri's work, turning techniques Montgomery introduced in 1973 into an unconditional bound; the pivotal move was treating zeros on and off the critical line as a single geometric object rather than two separate cases. Two mathematicians at Anthropic read and checked the paper, and the model also produced a machine-checkable version of the proof.
Related stories
Twenty tries at the same task cut the best agent's score by a third.2026-10-06A model learned to write without the method that trains every AI.
2026-10-06The US got 16 countries to put AI at the centre of science.
2026-10-05Mathematicians cracked five open problems using an ordinary chat box.
2026-10-05Untuned models solved agent tasks their polished versions could not.
2026-10-04