🔍 Read the full analysis: Does OpenAI’s AI Mathematics Have Somewhere To Go? 722 Proofs Raise The Question on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI published 722 mathematical manuscripts produced by an unnamed, unreleased model, covering 372 families of results selected from about 4,000 problems. The claims include answers to long-standing open questions, but OpenAI says they have not been confirmed by outside mathematicians; whether the work yields reusable ideas remains unknown.
OpenAI published 722 mathematical manuscripts on Monday, presenting results generated by an unnamed, unreleased model across 372 families of related problems. The work includes claims about several prominent open questions, but the company and outside researchers have not established that the results are correct; the larger question is whether mathematicians can verify and build on them.
OpenAI’s post and the associated GitHub repository describe manuscripts spanning number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. The company says the results came from roughly 4,000 problems posed to the model, then filtered by OpenAI for what it considered an appropriate level of significance. The average result used about three hours of ChatGPT Pro thinking compute, according to the source material.
The manuscripts are grouped into 372 families and published under the Apache-2.0 licence. Lean formalizations accompany many, but not all, results. OpenAI’s README cautions that some unformalized results could have issues. The source material says ten abridged reasoning summaries were provided, rather than summaries for all 372 families, and that the Riemann zero-free-region write-up was edited by people for readability.
The catalogue includes claims involving the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, isomorphism of nonabelian free group factors, a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12, the Hodge conjecture for CM abelian varieties, and conjectures in convex geometry. These are claims in model-produced manuscripts, not established solutions. OpenAI chief executive Sam Altman described the results as claims not yet confirmed by outside mathematicians, according to the source material.
722 proofs, one question: will any of OpenAI’s AI mathematics actually lead anywhere?
An unreleased, unnamed model produced claimed proofs of results that would each define a career. Sam Altman calls them “claims not yet confirmed by outside mathematicians.” The real question isn’t whether it’s impressive. It’s whether answers nobody understands become discoveries anyone can build on.
Same day: Alon, Bloom, Gowers, Litt, Sawin post a digested, human-verified version. The model for success.
Connes rigidity counterexample challenged within a day — constructed groups fail the required condition. Three rival machine “counterexamples” from different labs now circulate.
~10,000 agents, 88 hours, est. ~$22M at retail. Priority dispute; 25 Fields Medalists sign “A Severe Misalignment” — not saying it’s wrong, saying it’s not understood.
Altman now hedges at announcement — a shift from September. Verification has barely started.
Humans extract the technique, write it up, build on it. This is where downstream discovery comes from.
The question is answered; nobody learns anything reusable. Closes a door without opening a field.
The proof breaks, or proves a statement that doesn’t match the conjecture as mathematicians mean it.
The Unique Games Conjecture is the clearest case. Results like the optimality of Goemans–Williamson for Max-Cut are proved assuming UGC. A correct proof converts them all — no understanding required. A zero-free strip for zeta works the same way for prime-distribution results. Free group factors, Kadison, Mahler would redirect whole programmes — but how depends on the method, which means digestion.
Technology. A Navier–Stokes blow-up proof doesn’t change how anyone designs aircraft; engineering turbulence models never depended on the answer. Near-term consequences are mathematical, not industrial. “AI will cure cancer next” skips several steps.
“Verification abundance, adjudication scarcity” — making proof-checking cheap doesn’t reduce the burden of deciding what’s true and what matters. 722 manuscripts land on a review system built for a trickle, filtered by a selection nobody outside OpenAI made.
Humans re-deriving results, like Alon–Gowers et al. in May
Other people’s work building on these manuscripts
How many unformalized results survive expert checking
Do the Lean statements match the real conjectures?
Do any survive peer review?
Some of it, yes — where a literature is waiting (UGC), a correct proof pays off immediately; where a proof carries a new technique humans digest, it can open a field. Most of it, probably not on its own: at 722 manuscripts with 10 reasoning summaries, the Four Colour pattern is the likely default unless mathematicians are funded and given time. And some will be wrong — OpenAI says so itself. It’s an industry pattern, not one company’s: the forced-Euler result came from an Anthropic researcher, and rival machine-generated Connes “counterexamples” circulate from different labs. The proofs arrived this week. The discoveries, if they come, will arrive at the speed of human understanding.
Verification Will Shape the Payoff
The number of manuscripts and the fame of some problems make the release notable, but publication is not the same as mathematical validation. Each argument needs careful checking, and formal verification can help establish that a proof follows from its stated assumptions without automatically showing that the assumptions match the problem mathematicians intended to solve. OpenAI’s warning about unformalized work also means the degree of checking varies across the collection.
The more consequential test is whether a correct proof offers techniques other researchers can use. A result that resolves a major conjecture could alter related work, but a proof that is difficult to understand or depends on a long chain of case checks may settle a question without providing much new mathematical method. The source material frames the possible outcomes as work that people digest and extend, a correct but less generative solution, or an argument that fails or addresses a different statement.
The Unique Games Conjecture illustrates why verification could matter beyond one paper: a substantial body of theoretical computer science uses it as an assumption when establishing limits on approximation algorithms. If the relevant manuscript proves the conjecture as stated, researchers would need to revisit dependent results. That consequence remains conditional; the manuscript has not been independently confirmed.
mathematical problem solving software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Offer Caution
This is described as OpenAI’s fourth major mathematics release this year. In May, its model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians—Noga Alon, Thomas Bloom, Tim Gowers, Daniel Litt and Will Sawin—then posted what they called a digested, human-verified version. The episode offers one possible route from machine-generated work to material the field can assess: experts reconstruct, check and explain the argument.
An August release called “Ten Advances” included a claimed counterexample to Connes’s rigidity conjecture that was challenged within a day. A critique argued that the constructed groups did not meet the condition required by the conjecture. The source material also says multiple machine-generated counterexamples to that conjecture are in circulation, underscoring that a result’s headline description may not settle whether it addresses the precise mathematical claim.
In September, OpenAI announced a Lean-formalized proof concerning finite-time blow-up in the Navier–Stokes equations, produced with about 10,000 concurrent agents over 88 hours, according to the source material. That announcement prompted debate about priorities in AI mathematics. Twenty-five Fields Medalists signed a declaration titled “A Severe Misalignment of AI in Mathematics.” Their stated concern, as described in the source, was not that the proof was wrong, but that treating famous problems as benchmarks without human understanding could work against the aims of mathematics.
As an affiliate, we earn on qualifying purchases.
Which Proofs Will Hold Up
Independent verification remains the central unknown. The source material does not identify outside mathematicians who have validated the 722 manuscripts as a collection, nor does it report a confirmed status for each of the 372 families. Lean formalizations may help check some arguments, but the repository says not every result is formalized, and a formal proof still needs to correspond to the intended mathematical statement.
It is also unclear how OpenAI selected the problems and judged significance beyond its description of the filtering process. The model has not been named or released, and the source material gives no public account of how much human intervention went into the results apart from the stated exceptions, including human editing of the Riemann write-up. For now, claims about landmark solutions and their potential to reshape research should be treated as provisional.
As an affiliate, we earn on qualifying purchases.
Outside Review Sets the Timeline
The next step is independent scrutiny of individual manuscripts: mathematicians will need to inspect the arguments, test formalizations where available, and determine whether each result proves the stated claim. The source material does not provide a schedule for reviews or say when OpenAI expects assessments to be completed.
Readers should watch for detailed critiques, human-readable reconstructions and corrections to the repository. The release’s scientific impact will depend less on the headline count than on which arguments survive review and whether researchers can extract methods that lead to further work. Until those steps take place, the 722 manuscripts represent a substantial set of mathematical claims, not 722 confirmed discoveries.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
It published 722 mathematical manuscripts generated by an unnamed, unreleased model, organized into 372 families of related results.
Have outside mathematicians confirmed the results?
Not as a collection, according to the source material. Altman described the manuscripts as claims not yet confirmed by outside mathematicians, and the repository warns that some unformalized results could have issues.
What is the Unique Games Conjecture claim?
One manuscript claims a proof of the Unique Games Conjecture, an open problem connected to limits on approximation algorithms. The claim remains unverified in the supplied source material.
Why does a correct proof need to be understandable?
Understanding a proof can reveal methods that researchers apply elsewhere. A proof may settle a question while producing little reusable theory if its reasoning cannot be readily understood or extended.
What happens next?
Mathematicians must assess the arguments and formalizations. No review timetable or final verdict for the manuscripts is provided in the source material.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
