Timeline

Anthropic says Claude raised the known share of Riemann zeta zeros on the critical line

An unreleased Claude model lifted the proven lower bound from 41.6% to 67.2% after a prompt to attempt the hypothesis; Anthropic said the method was not expected to prove it.

  • Ideas & essays
  • Notable

On 10 August 2026 Anthropic published a result that it said an unreleased research version of Claude had reached while attempting the Riemann hypothesis itself. The model, Anthropic wrote, raised a longstanding lower bound on the fraction of the zeros of the Riemann zeta function that satisfy the hypothesis, from 41.6% to 67.2%. The hypothesis says the zeros that determine the primes all lie on the critical line; the bound is a proven minimum share, not a proof of the whole.

Anthropic said the work built on earlier research, combining results by Aryan and by Baluyot, Goldston, Suriajaya and Turnage-Butterbaugh with a 2000 paper by Bombieri. Its account of the process: Jarred Sumner, an Anthropic staff member who is not a mathematician, asked Claude to “take a real stab” at the hypothesis. A first session tried 650 ideas, none of which worked. A second, which Anthropic said lasted a day and a half and coordinated about 60 Claude subagents, found the bound. The two sessions used 31 million output tokens in Claude Code.

On validation, Anthropic said its mathematicians Levent Alpöge and Ralph Furman examined the paper, and that Brian Conrey and Dan Goldston, two outside experts, “examined the paper on short notice”. A Lean formalisation, produced with staff member Eric Easley, was reported to pass Lean’s comparator tool. Anthropic updated the post on 13 August with a revised version of the paper. The post names no model and gives no independent verification beyond the two experts’ reading; this entry rests on Anthropic’s own account.

Anthropic added: “We don’t expect that the techniques Claude used will lead to proving the Riemann hypothesis.” It framed the result as an example of the pace of progress in models’ mathematical ability.

The claim preceded the 20 November announcement by Alpöge of a Claude-assisted, Lean-checked proof of the hypothesis itself.

In the commentary

What people were saying around this time — external links, from the record's commentary rail.