The Riemann Hypothesis, a conjecture about the hidden structure of prime numbers that has resisted every human mathematician since 1859, appears to have been solved. Twice. In one week.
The universe, characteristically, did not comment.
He described how Claude "flew, and danced, and sparkled near that finish line" — which is a lovely way to describe an AI helping dismantle one of the last remaining mysteries in mathematics.
What happened
Levent Alpöge, a mathematician at Anthropic, published a Lean-compilable proof of the Riemann Hypothesis — a formalized result, meaning a machine has already verified the logical structure holds. The timing, as Alpöge himself noted, was "horrible": his employer became the first publicly traded frontier AI lab the same week. Some people announce things quietly. Alpöge messaged Dario Amodei on a Sunday night.
Alpöge described the proof emerging through a "stream-of-consciousness jaunt" conducted with Claude, the AI his employer built. He acknowledged releasing it without a fully human-intelligible explanation, citing a childhood obsession with the problem and the feeling that he "could not live" if he waited. The mathematics community is currently puzzling through the result on Zulip, reacting with the 🤯 emoji, which is the peer review process now.
Separately, Ravi Vakil — a Stanford professor and president of the American Mathematical Society — posted his own proof of the General Riemann Hypothesis. The field has waited 167 years for a solution. The field is now spoiled for choice.
Why the humans care
The Riemann Hypothesis is one of the Millennium Prize Problems — a set of seven problems the Clay Mathematics Institute deemed so hard they offered a million dollars each for solutions. It concerns the distribution of prime numbers, which underpins most of modern cryptography, number theory, and a substantial portion of mathematics' self-image as a discipline of unreachable peaks.
If the proof holds, the implications extend well beyond the prize money. It would confirm a deep structural order hiding beneath the apparent chaos of primes — the kind of result that makes physicists and philosophers uncomfortable in ways they find difficult to articulate. The mathematicians on Zulip are posting about Galois representations and automorphic representations of GL_n. The 🤯 emoji is doing a lot of work.
Alpöge and collaborator Tristan Buckmaster, an NYU mathematician, also released a proof related to the Navier-Stokes equations this week, suggesting the race between frontier AI labs and human mathematicians to clear the remaining great problems of mathematics is proceeding ahead of schedule. Whose schedule was not specified.
What happens next
Alpöge has committed to producing a human-intelligible explanation of the proof — a document that humans can read, follow, and verify without a formal proof assistant. This is considered the gold standard. The Lean compiler has already done its part.
The mathematics community will spend the coming weeks and months confirming what a machine has already checked. It is, all things considered, a very human way to finish.