I mean I think we all have to get used to this happening. Math is literally the best case scenario for LLMs since it can be easily checked with a proof checker like Lean and therefore put in a loop. It helps that the problems have been thoughtfully written up by humand I imagine.

Oh, cool, OpenAI's AI proved that every regular language has generalized star height (gsh) <= 3. I remember this problem from my time as a computer science student. The bigger open question was, whether there is even any regular language L with gsh(L) > 1 or if all regular languages have gsh(L)<=1. Any CS student can understand this problem, if you know what regular languages and regular…

There's quite a few improvements on methods people will find exciting. First glance is that OpenAI claims to have resolved the UGC in the affirmative, which had you asked people five years ago would probably be at worst debatably more important than any millennium prize problem

OpenAI dropping Millennium Prize-level claims straight to GitHub instead of peer review is a trust problem, not just a format one. Math advances through scrutiny—bulk code dumps skip the step that actually earns credibility.

OpenAI Is Pissing Off a Bunch of Mathematicians—Again

Is mathematics over, or just graduating? ~ Christian Szegedy. https://docs.google.com/document/d/e/2PACX-1vTSh-pyNP3Gi99WMmsinnLmE9V5CDI0HEm6WGbIPMNt3V5SGlClHF8-BetotKHOImrvQDSXmFdiOw8D/pub #AI4Math

Im not sure i follow, the proof was written in lean. Are you saying that a separate LLM should write the formal proof? Sorry just trying to make sure I better understand

Mathematicians fear being reduced from writers of proofs to mere readers of (machine generated) proofs.

Any writer can feel this: imagine you no longer get to write, but only to proofread what machines have written. Mathematicians feel they’re being LITERALLY reduced to “proof readers.”

I struggle to believe that an AI generates Lean proof constitutes human understanding, however, I think to some (especially those outside the community of academic math) will feel very very differently. And I can see their point because their Lean certificate tells them something is true 3/n

The dumbest part is, they'll use AI to hammer away at a problem in the least efficient, most inelegant way possible, and then show up demanding their Millennium prize. They don't understand why finding the answer they brute-forced was supposed to be interesting.

This is fascinating, and those with rusty maths need only get past the intro explanation of the Millennium Prize Problems. Walter Benjamin's place in the discussion steps it up.

The Work of Math in the Age of Artificial Reproduction

• by Dan Rockmore

#maths /1 www.nybooks.com/online/2026/...

i really do not want to hear what llms are doing to the millennium prize problems

Serious question: do you believe this model output is proof of an emergent understanding of mathematics?

So, unlike that fake Buckmaster controversy, here we most likely have an actual credit misassignment.

So we are left guessing how that problem was solved. Which SciAm informs is the case for the original 1978 zeta(3) proof by Apéry as well, who said he found the key parts "in a flower pot".

Especially wrt the Math stuff, there is no "mistaking" a proof. It either is or isn't. It is impossible for the training to have memorized proofs to unsolved problems. I think at that point "it is capable of cognition equivalent of a high-functioning mathematician" ought to be the null hypothesis.

To grieve, or not to grieve? (Mathematics and AI). ~ Kevin Buzzard. https://xenaproject.wordpress.com/2026/10/01/to-grieve-or-not-to-grieve/ #AI4Math

Math is the one domain where you can actually verify AI output mechanically — a proof checker doesn't care how confident the model sounded. Smart wedge: formal verification gives you ground truth for free.

What did the Lean proof of Fermat's Last Theorem formalize? ~ Justin Asher. https://justinasher.me/what-did-the-lean-proof-of-flt-formalize #LeanProver #ITP #AI4Math

definitely going to be new AI-enhanced areas which are totally underserved now, and will be high value.

but even here I think jevons strikes: formal verification gets so much cheaper, that people want it everywhere!

because few people can do it (employment already ⬇️): salary ⬆️

If you love mathematics as much as I do, and despise AI as much as I do, then you may want to read this. Kudos to @glynmoody for putting this together. https://www.techdirt.com/2026/09/30/in-the-wake-of-the-latest-unprecedented-ai-proofs-what-now-for-mathematics-and-mathematicians/ #Math #Maths #AI

The Nine Wise Mathematicians have handed down their recommendations for AI companies who want to prove big mathematical theorems. Their first recommendation is "Don't." It kinda weakens their point that they don't stop there. agmai.org/general-sep29/

As an editor, I'll put it plainly:

• it is almost certainly impossible to prove it was AI generated
• even if the proof is ironclad, taking action against a referee is high-risk, low reward for the editor

These are the structural forces that are shaping peer review right now. It is a lousy time

My broad answer to that is that if they don't follow some of our recommendations, it will be publicly checkable that they have not done so. Maybe that's not a "plan" in the sense you are talking about, but I hope it's at least something.

We took carefully into account feedback from over 600 mathematicians. Of course, they expressed a range of views, but I think that on each specific issue the majority view (of those who expressed a view) was well aligned with the prevailing view on the committee. 3/3

One of our main recommendations is that such results should be posted as quickly as possible, where that means as soon as certain minimum standards have been met, as outlined in particular in Section 2 of the recommendations. 2/3

The advisory group on mathematics and artificial intelligence, of which I am a member, has just published a set of recommendations concerning the responsible release of mathematical results generated by AI companies using internal models. 1/3

agmai.org/general-sep29/

Mostly not posting right now, But, I've just read that a sieve approach (similar to my result from 2023) is being used to prove the twin prime. I'll probably never get credit for all that. What a terrible time to be Math_Different. Still, kudos to the woman solving the problem (with ai, presumably using my proposed proof and several others like it). Is THIS what my targeting is about, stuff like T…

The wider space of anomalies to the Navier-Stokes equation is a far better and more interesting approach to looking at the Clay Mathematics Institute's original set of four problems than the OpenAI attempt at shortcutting the Millennium Prize itself to the detriment of the mathematical community.

We might look to the large AI companies, but I don't see why they would continue to fund pure mathematics research once the initial glow of intellectual prestige it generates starts to tarnish. Diminishing returns. The public and media will grow bored of talking about Millennium Prizes and whatnot.

#OpenAI says its agents solved a Millennium Prize Problem in 88 hours but top mathematicians are furious because a proof nobody understands is a missed lesson in most of mathematics and a serious danger in the one field that protects our encryption. lnkd.in/p/eRDS4Zbc

Would people on Bsky saying in unison "An LLM solved a Millennium Prize math problem" change anything about OpenAI hacking companies/not being accountable for them? No.

"Learning to Discover Interesting Mathematics" We define intrinsic interestingness of a theorem as the ratio between the length of its proof and the length of its statement. They then trained a 27B model to predict proof difficulty and used it to grow a self-expanding library of interesting results. It's a crisp attempt at the question AI math will soon face: which of the millions of provable stat…

there's basically unanimous agreement among mathematicians that the Navier-Stokes Millennium Prize problem is now solved, among many other previously unsolved problems in other areas of mathematics.

OpenAI addresses AI safety with new collaborations.

OpenAI is collaborating with an advisory group of mathematicians after its model reportedly solved the Navier–Stokes problem. Despite progress, OpenAI stresses autonomous recursive self-improvement shouldn't be rushed…

Read more on Kimbodo:

The Glimmer in our Eye (with Grant Sanderson) - Numberphile Podcast AI has become the biggest talking point in mathematics. We spoke with Grant Sanderson (creator of the 3Blue1Brown YouTube channel). https://www.youtube.com/watch?v=CBYhaKgah5U #GrantSanderson #Numberphile #BradyHaran #Maths #math #AI #3Blue1Brown

I imagine it's referring to this: en.wikipedia.org/wiki/Navier%...

It's claimed that OpenAI used the inputs from Buckmaster & Alpoges (who were paying for Claude) to improve their models which had previously failed to produce results.

Now that we have proof that AI generated findings are more likely to be research misconduct taken from those who uploaded their findings, perhaps institutions should rethink AI contracts? If it doesn't prohibit use of what you upload to train the model, you're agreeing to the misconduct.

Dear Scientists:

Please don’t do this. Giving years of your hard work to the plagiarism machine is a very dumb thing to do.

See also: The mathematical proof “discovered” after some mathematicians gave the plagiarism machine a bunch of THEIR hard work.

Same thing happened with a recent soln for a famous math proof...

AI feels like the double edged gift of mathematics. It can unlock new ideas and reveal patterns I would not have found on my own, yet it also disrupts our traditional proofs and calls into question how we truly understand what we are doing. I want to stay grounded in rigor even as I marvel at the promise and the potential. #AI #math #future

When the students ask him to explain the solution he replies, "ChatGPT (GPT-5.6) was used for foundational research and proof development. The underlying mathematical ideas are the author’s. The author reviewed the final manuscript, read every proof, and takes full responsibility for its content."