FoxTPNL / Wikimedia Commons (CC BY 4.0) After claiming to solve one Millennium Prize problem with 10,000 AI agents, OpenAI hints at cracking a second while over 1,000 mathematicians cry foul
OpenAI appears to be closing in on the Hodge Conjecture, one of the most notoriously abstract unsolved problems in mathematics, but the company is treading carefully this time. After its September announcement of solving the Navier-Stokes existence and smoothness problem sparked a full-blown revolt in the math community, the AI lab is reportedly taking a more measured approach to disclosing its next breakthrough.
The company has publicly acknowledged making “substantial progress” on a second, unnamed Millennium Prize problem. It hasn’t confirmed which one. But speculation has converged heavily on the Hodge Conjecture, a problem so esoteric that even explaining what it asks requires a graduate-level understanding of algebraic geometry and topology.
The Navier-Stokes precedent #
To understand why OpenAI is walking on eggshells, rewind to September 9-10, 2026. That’s when the company announced it had produced a solution to the Navier-Stokes existence and smoothness problem, one of seven Millennium Prize Problems that the Clay Mathematics Institute designated in 2000, each carrying a $1 million bounty.
The method was as audacious as the claim. Approximately 10,000 AI agents collaborated to produce a 166-page research paper, complete with formal verification code written in Lean, a proof assistant language that mathematicians use to check logical rigor. The entire process took 88 hours.
The paper is currently under review by the Clay Mathematics Institute, which has historically taken years to verify claimed solutions to its prize problems. Grigori Perelman’s proof of the Poincare Conjecture, the only Millennium problem solved to date, took roughly four years of community verification before the Clay Institute officially recognized it in 2010.
The math community fires back #
The backlash was swift and coordinated. More than 1,000 mathematicians signed an open letter criticizing AI’s expanding role in mathematical research, going so far as to label OpenAI’s approach “research misconduct.”
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Then came the heavier artillery. A group of 25 Fields medalists issued a joint statement expressing concerns about what they called “severe misalignment” between AI developments and mathematical ethics.
OpenAI responded to the controversy by withdrawing its sponsorship from a math hackathon at Caltech.
Why the Hodge Conjecture matters #
The Hodge Conjecture, proposed by Scottish mathematician William Hodge in 1950, asks whether certain geometric structures on complex algebraic varieties can be built from simpler, well-understood pieces called algebraic cycles. It sits at the intersection of algebraic geometry, topology, and complex analysis.
The careful rollout #
OpenAI’s decision to slow-walk its next announcement reflects a company learning from experience. Mathematics doesn’t move at product launch speed. Proofs are typically circulated as preprints, discussed at seminars, picked apart by specialists, and gradually absorbed into the field’s body of knowledge over months or years. OpenAI essentially tried to speedrun that process with a press announcement and a Lean codebase, and the community pushed back hard.
The Clay Mathematics Institute’s review of the Navier-Stokes paper will be the first real test. If the proof holds, it fundamentally changes the conversation about AI in pure research. If it doesn’t, OpenAI’s next announcement, Hodge Conjecture or otherwise, will face an even steeper credibility hill to climb.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our