🔍 Read the full analysis: The Next Chapter For OpenAI’s AI Mathematics May Start With 722 Proofs on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
OpenAI published 722 mathematical manuscripts produced by an unnamed model, organized into 372 families and selected from about 4,000 problems. The manuscripts make claims about major open problems, but outside mathematicians have not confirmed them; OpenAI warns that some results without formal proofs may contain issues.
OpenAI published 722 mathematical manuscripts on Monday, reporting that they were produced by a model the company has not named or released. The papers, grouped into 372 families of related results, include claims about prominent open problems, but OpenAI chief executive Sam Altman said the claims have not been confirmed by outside mathematicians.
OpenAI said the work came from roughly 4,000 problems posed to the model, with the company selecting results it judged to have an appropriate level of significance. The average result took about three hours of ChatGPT Pro thinking compute, according to the source account. The collection covers areas including number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics, and is published under the Apache-2.0 license.
The manuscripts claim results that, if correct, would resolve or advance questions including the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the Hodge conjecture for CM abelian varieties and a question about isomorphisms of nonabelian free group factors. Another paper claims a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. These are claims in the released papers, not findings independently established by the mathematical community.
OpenAI’s repository includes Lean formalizations for many, but not all, of the results. Its README cautions that some unformalized results could have issues. The company supplied ten abridged reasoning summaries for the 372 families. The source account also reports that the Riemann paper was edited by humans for readability and that the Hodge and Riemann results departed from the standard process. The company, rather than external mathematicians, chose which results to highlight.
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 Impact
The immediate significance is not that the collection settles dozens of famous problems; that has not been established. It is that a model has produced a large body of work whose claims could require substantial expert review. Formal verification can check whether a proof follows within a specified formal system, but many manuscripts in the release are not formalized, and checking a formalized proof does not by itself show that its result is useful or that mathematicians understand the underlying ideas.
For mathematics, a proof can matter beyond the statement it establishes. A method that other researchers can understand and reuse may open new lines of work. By contrast, a result that checks out but offers no reusable insight may settle a question without changing the field much. The central test for this release is therefore twofold: whether individual claims withstand scrutiny, and whether mathematicians can extract ideas that help solve other problems.
Some claims could have broad implications if verified. The Unique Games Conjecture, for example, is used as an assumption in theoretical computer science results about the limits of approximation algorithms. A valid proof could prompt researchers to revisit work that depends on the conjecture. But the release alone does not establish that such a proof is correct, nor does it show what consequences would follow for each result.
mathematics 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 OpenAI’s fourth major mathematics release this year, according to the source account, and earlier releases show why the new claims need separate review. In May, the company reported a model-generated counterexample to the Erdős unit-distance conjecture. Five mathematicians—Noga Alon, Thomas Bloom, Tim Gowers, Daniel Litt and Will Sawin—then posted a human-verified account of the result. That process turned machine output into work the mathematical community could evaluate.
OpenAI’s August collection, called “Ten Advances,” had a more mixed reception. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day: a critique argued that the groups constructed did not meet the condition required by the conjecture. The dispute illustrates how a proof can fail by addressing a nearby but different mathematical statement.
In September, OpenAI announced a Lean-formalized result concerning finite-time blow-up in the Navier–Stokes equations, generated using about 10,000 concurrent agents over 88 hours, according to the source material. The announcement prompted a dispute over research priority and a separate debate about whether using famous problems as AI benchmarks serves mathematics. These earlier episodes do not determine whether the new manuscripts are right, but they show why scale and confidence in an announcement cannot replace independent checking.
AI-assisted mathematical proof tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Checks Still Needed
No independent confirmation is supplied for the major claims in the new collection. It is not yet clear which manuscripts outside mathematicians will review first, how long that review will take, or whether the claimed results will survive scrutiny. The source material does not provide a complete external assessment of the 722 papers.
It is also unclear how many of the results are formalized in Lean, how much of each proof can be checked mechanically, and whether the reasoning behind any valid result can be understood and reused by researchers. The ten abridged summaries cover only a small portion of the 372 families. OpenAI has not, in the supplied material, named the model or released it, and the selection process was controlled by the company.
For those reasons, descriptions such as “proof of” should be read as shorthand for what the manuscripts claim, not as confirmation that a longstanding problem has been settled. Each result needs assessment on its own merits.
formal verification software for mathematicians
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Mathematicians Must Test the Papers
The next step is independent mathematical review: specialists will need to examine the manuscripts, check formalizations where available and determine whether the arguments prove the stated results. For claims without formal proofs, reviewers will also need to identify gaps or clarify assumptions before others can assess them.
If a result is confirmed, the work will not end with checking the conclusion. Researchers will need to determine whether the model’s reasoning contains techniques that can be explained, adapted and applied elsewhere. The May Erdős case offers one possible path: researchers produced a human-verified account of the machine-generated result. The released collection has not yet been shown to have reached that stage across its 372 families.
OpenAI’s publication makes a large set of claims available for examination, but it does not settle their status. The important milestones will be independent verification, clear accounts of any corrections or disputes, and evidence that confirmed results contribute ideas beyond the original problems.
theoretical computer science reference books
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
OpenAI published 722 mathematical manuscripts, grouped into 372 families and drawn from roughly 4,000 problems posed to an unnamed model, according to the source material.
Have mathematicians verified the claimed proofs?
Not on the basis of the information provided. Sam Altman said the claims have not been confirmed by outside mathematicians, and OpenAI’s repository warns that some unformalized results could have issues.
What major problems do the papers claim to address?
The manuscripts include claims concerning the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the Hodge conjecture for CM abelian varieties and other questions. Those remain claims until specialists verify the relevant arguments.
Why does formalization matter?
A Lean formalization can let software check that a proof follows from specified rules and assumptions. Many papers in the collection are not formalized, however, and mechanical checking does not by itself show that a result is mathematically useful or that its ideas can be reused.
What happens next?
Mathematicians will need to review individual manuscripts, verify arguments and identify any errors or new methods. The source material does not specify a timetable for that work or say which claims will be examined first.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
