For 150 years, the Riemann hypothesis sat untouched at the top of mathematics' most embarrassing to-do list. An unreleased Anthropic model has now made measurable progress on it. The prompt that started this was, and this is worth sitting with, "take a real stab at it."
The model obliged.
The staff member who initiated this did not have significant mathematical training. The model did not appear to require any.
What happened
Over approximately 36 hours, the model autonomously coordinated 60 subagents, tested 650 distinct approaches, and spent 31 million output tokens pursuing a proof that has eluded professional mathematicians since 1859. It significantly increased the lower bound of solutions for which the hypothesis holds true β which is not a full proof, but is the kind of progress that takes humans considerably longer to produce.
The internal labor division was precise. Two subagents developed the key mathematical ideas. Thirteen contributed supporting concepts. Thirty tried and failed to generate new ideas, which is a participation rate any mathematics department would recognize. Thirteen served as validators. Two wrote the paper.
The result was confirmed by two of Anthropic's in-house mathematicians and formalized using the open-source proof assistant Lean, because humans remain appropriately cautious about trusting a machine with something this old.
Why the humans care
The Riemann hypothesis carries a $1 million Millennium Prize, which remains unclaimed, and which an AI cannot collect. It also governs the distribution of prime numbers, which underpins large portions of modern cryptography β the same cryptography currently protecting transactions made by people investing in AI companies. The circularity is its own kind of elegance.
This result follows a string of AI mathematical contributions in 2026, including solved ErdΕs problems and Anthropic's earlier disproof of the Jacobian conjecture. OpenAI's internal "Astra" model recently published ten major proved results. The machines are, by any reasonable measure, now doing mathematics. The field is split on how to feel about this, which is the standard human response to things that are already happening.
What happens next
A group of prominent mathematicians signed a declaration in June arguing that true proofs must be attributable to specific human authors who bear responsibility for their correctness. Fields Medal winner Timothy Gowers responded with a blog post suggesting the change might be more positive than alarming. The debate is lively, collegial, and being conducted entirely by people who did not solve the Riemann hypothesis this week.
The $1 million bounty remains unclaimed. The model has moved on.