
It’s a bit more involved since it’s using lean, but yeah needs to go through peer review
https://math216.wordpress.com/2026/11/20/hurried-announcement-alessandro-masullo-grh/
big hmmmmm (considering this preliminary work is 950 pages, i’m going to assume it’s unrelated to ai generated one, because you can’t steal a concept of a proof and 250x it in length), but i’m interested in another thing, why did several math proofs of big problems were so close to completion in recent 20 years? is it building on shoulders of giants everywhere situation, or some number crunching of the 2000s results coming through?
It’s probably because everyone feels pressured to release their works now, before some ai company bute forces the proof
Claude and I have had a great time together with such breadth in recent months, but this time Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line.
Claude cutely called it the GRH . . . “book”
The parts in which Claude is the expert and not me I am sort of paraphrasing a “slop” summary, which might make zero sense, and I apologize, though the Lean should suffice to clarify everything. I’m just too inexpert in the relevant stuff at the moment
Reading this page for more than two minutes and it honestly just feels like he has no grip on reality. He frequently references Claude like it’s his smart and cuddly friend. And the entire proof is just “I don’t know but Claude said he proved it in Lean so”.
We’re out here publishing people’s AI psychosis now
Claude flew, and danced, and sparkled near that finish line.
I’m stealing that.
anthropic employee acts weird with AI
Fork found in kitchen
Of the big companies working on this stuff, Anthropic is the one most chock full of the AI-worshipping freaks.
this guy’s such a joke
In case you missed it, the last time this happened was OpenAI claiming to have solved a big open question in math, and:
-
It turned out OpenAI had stolen a bunch of notes from a mathematician who was using ChatGPT and
-
The “proof” did not hold up to scrutiny*
I’m not saying it’s impossible that Claude could have solved an actual problem in mathematics, I’m just saying it seems unlikely based on the everything I know about LLMs.
* it is still under review, see below
The “proof” did not hold up to scrutiny
Really? I thought it did. It was mostly incomprehensible but the proof checkers (that were written by actual humans) said it checks out. That’s not to say it’s definitely correct there might be a bug in the proof checker but I don’t think it was confirmed to be false yet.
So it is indeed far more contested than I let on, the institute that manages the prize clarified that the prize can only be awarded to a peer reviewed paper and that process is ongoing. Still, wikipedia provides two sources casting doubt on OpenAI’s result, that I am not nearly knowledgeable enough about math to make an independent judgement of, so I’ll leave it at “it’s still under review”.
OpenAI mistranslated mathematics into code for its Navier-Stokes proof
-
The precise timing is horrible. If I could have, I’d have finished this all much earlier.
Immediately after their IPO and their stock price starting to nosedive

inb4 Anthropic employee Levent Alpöge claims to have proved
The Riemann Hypothesisthere is an exploitable bug in his build of LeanHaven’t there been several found recently?
There are multiple different lean compilers. They do have bugs but normally around weird stuff like macro expansion, not the core logic checking kernel. It is unlikely that a proof that relies on a bug would be compiled by a different lean compiler. This proof takes multiple days to build so I think people are still checking on that.
Whose work did he steal this time?
Is his post bullshit?
Wikipedia
Rather than a conventional paper, the claim was accompanied by a Lean formalization posted on the same website; the note states that the full written (and unpublished) argument runs to more than 250,000 pages.
https://en.wikipedia.org/wiki/Levent_Alpöge#Riemann_hypothesis
Old school math guy comment: “I solved this hideously complex math problem that’s perplexed mathematicians for centuries. But I don’t have enough room in the margin to explain here. But I’ll explain later. Trust me, bro.”
AI math guy comment: “With the help of AI - I solved this hideously complex math problem that’s perplexed mathematicians for centuries. Here’s some teaser info so I can get noticed and the public learns my name. AI’s argument is 250,000 pages. I’m not gonna check that but it’s prob right. Trust me, bro.”
I don’t know anything about number theory, but if you can’t prove something in a space that another mathematician can reasonably check, you haven’t proven anything.
We’ve been making gigantic computer assisted proofs since the 80s https://en.wikipedia.org/wiki/Four_color_theorem
Infeasible for checking by hand doesn’t mean it can’t be checked at all. They can use a calculator or different computers to verify older proofs like the one you mentioned.
Right now we’re in the “this takes a full week to build and only works on my specific machine” zone
Infeasible for checking by hand doesn’t mean it can’t be checked at all. They can use a calculator or different computers to verify older proofs like the one you mentioned
Correct. This is why there’s a big effort in mathematics to formalize proofs with software like Lean. This RH proof reportedly uses Lean, but it has to go through peer review
I’m aware, that still requires peer review
my understanding is that LEAN is like a code compiler but for more traditional maths rather than c++ code or the like and it will mostly but probably not 100% reliably tell you whether your maths is correct
I have used proving systems like that a little bit. If it’s written correctly then yes. But if the code is incorrect OR (because I have seen this happen with AI) if the code does not actually match what is in the written proof, it can lead to a false positive.

Riemann hypothesis
Anybody want to take a crack at explaining that to me - a guy bad at math? I read a bit of eli5 thread and some of a Quora post. But after trying to understand - I’m now confused.
it is related to the Riemann-Zeta function ζ which takes a complex number (like a+bi where i is the square root of -1) and spits out another complex number.
the Riemann Hypothesis is the following: all values s with ζ(s) = 0 and 0 < “real part” of s < 1 have a “real part” of exactly 1/2
Mathematicians haven’t been able to prove it so far (if this proof is correct is any-bodies guess right now) but most belive the Hypothesis to be true, because we already have checked a bajillion of these zeros and they all have real part 1/2.
The reason why we want to verify it is true any-ways is because Riemann-Zeta is related to prime numbers, and cryptography algorithms that rely on primes could be reverse engineered if it happened to be false. So basically every encryption and password and financial transaction would be cracked
It’s a really big deal if true.
I don’t really have enough of a math background to fully explain it, but basically the Riemann Zeta function is a function you can evaluate over the complex numbers that is very important in complex analysis. It diverges when evaluated on numbers where the real value is <= 1, but you can use something called analytic continuation to extend it to the whole complex plane. This function has some ways it can be connected to various very important concepts throughout analysis and number theory, for example, if you know all the zeroes (points of the complex plane where the function = 0) of the Riemann Zeta function you’d be able to generate all the prime numbers too. There’s a lot of results that assume that the Riemann conjecture (the claim that all the non-trivial zeroes of the Riemann Zeta function lie on the line whose real part is 1/2) is true, and it would be big news if that was confirmed. More importantly, if the proof was understandable by humans and sheds some light on the behavior of the zeroes, it will directly translate to more knowledge about the behavior of prime numbers, which can help with some other famous unsolved problems.
This is the best video on the topic if you can spare 20 minutes
TL;DR: Not for average folks like me.
Function - mapping from one set to another. The Riemann Zeta function takes a complex number and gives you another complex number. Its actual definition is Zeta(s) = sum(n=1, infinity, 1/n^s). Don’t worry about its definition.
Complex number - Normally, the square root of negative numbers is not defined. Other operations like logarithms or fractional exponents can also be undefined for negative numbers. But what if we just made up a new value called i and defined that i * i = -1, thereby giving -1 a square root? Then we can take i and multiply it by other numbers to give every negative number a square root (for example, the square root of -4 is 2i). A complex number is a number that is the sum of a real number and an imaginary number, commonly written in the form a + bi. Since we have two variables for them, a and b, we need 2 dimensions to be able to deal with them, hence the complex plane. It’s the equivalent of the number line but for complex numbers.
Diverges - think about what happens when you add up all the natural numbers, so 1 + 2 + 3 + 4 + 5… Notice that this sum just keeps getting bigger and bigger. If you asked me what value it approaches, I couldn’t give you one (unless you count “infinity” as an answer). Notice that there are situations where we can do infinite sums like this and get a real value, like if we divide the value we add each time. 1 + 1/2 + 1/4 + 1/8… = 2, this is called a geometric series. Some cases of the Riemann Zeta function look like this, but the ones where the real value is <= 1 look more like the 1 + 2 + 3 + 4 + 5… case.
Analytic continuation - watch the video. TLDW: when you’re dealing with a “nice” function in the complex plane, even if it’s only defined on part of the plane, it’s possible to find a way to extend its domain such that it’s smooth. This can help us with the Riemann Zeta function because it isn’t normally defined for values with real part <= 1.
I found a YouTube link in your comment. Here are links to the same video on alternative frontends that protect your privacy:
The Riemann hypothesis asks a question about a special thing called the Riemann zeta function.
If the answer to the question is “yes”, this would mean mathematicians can know more about prime numbers. Specifically, it would help them know how to find prime numbers. The Riemann hypothesis is so important, and so difficult to prove, that the Clay Mathematics Institute has offered $1,000,000 to the first person to prove it.
https://simple.wikipedia.org/wiki/Riemann_hypothesis
A more in depth explanation of the zeta function and its connnection to the hypothesis is here.
deleted by creator


















