AI and Human Proofs of Riemann Hypothesis Posted Same Week

Anthropic mathematician Levent Alpöge posted a Lean-compilable proof of the 1859 Riemann hypothesis on Sunday, the same week his employer became the first publicly traded frontier AI lab, while Stanford professor Ravi Vakil circulated a separate, human-written draft of the General Riemann hypothesis by postdoc Alessandro Masullo, TechCrunch reported.

Two proofs, one chaotic week

Alpöge announced the result in a personal essay describing a “stream-of-consciousness jaunt” through the argument with Anthropic’s model Claude, which he said “flew, and danced, and sparkled near that finish line.”

He messaged CEO Dario Amodei Sunday night asking how to handle the announcement. “The precise timing is horrible,” Alpöge wrote. Within days, independent researchers including MIT mathematician Andrew Sutherland confirmed the Lean code compiles successfully, Orda Global reported.

Masullo’s manuscript, the product of twelve years’ work, arrived almost simultaneously. Vakil, president of the American Mathematical Association, posted the draft on Zulip with a note that it was “all human, with no AI.” The original plan had been to polish the writing for a year before release. “Now it seems best to put out the current draft,” Vakil wrote.

GettyImages
Photo: ordaglobal.com

What the Lean verification actually proves

Lean is an open-source proof assistant that translates mathematical statements into executable code. When a proof compiles, every logical step has been checked by the kernel — no gaps, no hand-waving. Alpöge’s repository passes that test.

Chris Birkbeck, a number theorist at the University of East Anglia, inspected the Zulip thread and described the core as an automorphy-lifting argument relying on Galois representations attached to regular algebraic, polarized cuspidal automorphic representations of GLn over CM fields, together with base change, Jacquet–Shalika, Clozel’s purity, Bernstein–Zelevinsky, the main theorem of complex multiplication, and Deligne–Ribet.

“For my part, I tapped out of mathematics at college calculus,” the TechCrunch author admitted, “but that comment was followed by 20 readers reacting 🤯.”

Compilation does not equal human comprehension. Alpöge acknowledged releasing the work without a “truly human-intelligible explanation,” an omission he called guilty but unavoidable: his childhood obsession with the problem meant he “could not live” if he delayed. He and NYU collaborator Tristan Buckmaster also dropped a separate Navier–Stokes proof this summer, part of a broader race between frontier labs and mathematicians to resolve open problems.

The interpretability gap

The community now faces a novel bottleneck: a machine-verified proof that few humans can yet read, teach, or extend. “What to watch next is the gap between machine-checkable and human-understandable,” GlassEpoch observed. Alpöge has committed to producing a narrative exposition. Until then, the result sits in an unusual state — formally verified, publicly contested, and largely unexplained.

Masullo’s draft carries the opposite problem: human prose that has not yet survived mechanical scrutiny. “If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere,” Masullo wrote in his introduction.

AI claimed a solution to the 167 year-old Riemann Hypothesis — and a human did, too
Photo: glassepoch.com

Millennium Prize and the road ahead

The Riemann hypothesis is one of seven Clay Mathematics Institute Millennium Prize problems, each carrying a $1 million award. The conjecture asserts that all non-trivial zeros of the zeta function lie on the critical line Re(s) = 1/2, imposing a hidden order on the distribution of prime numbers.

“The cohomology bound is proved by what looks like an automorphy-lifting argument. It genuinely uses the existence of Galois representations attached to regular algebraic, polarized cuspidal automorphic representations of GL_n over CM fields, along with base change, Jacquet–Shalika, Clozel’s purity, Bernstein–Zelevinsky, the main theorem of CM, Deligne–Ribet and more.”

— Chris Birkbeck, number theorist, University of East Anglia, on Zulip

Masullo’s manuscript is circulating in PDF form.

Photo of author

Technology Desk

Technology Desk at Archyde.

Ricardo Quintela Receives First Experimental Cherry Harvest in La Rioja