It's not all that surprising. As human knowledge advances, making the last leap becomes easier and easier, but this is not apparent until that leap is actually made. When it is made, it is not at all unusual for more than one person to make it at more or less the same time. The classic example is Newton and Leibniz both inventing calculus at the same time.
It's not surprising because the other person has been sitting on the proof for a year, and intended to sit on it for another year - and for very reasonable reasons.
So for two proofs to be announced on the same day they only had to overlap in this two year period.
Makes sense if you think about it. Lots of people maybe are 99% but are waiting on polish and refinement or lack confidence. Then, one gets published and pushes others who are holding out to publish.
I think there's another bit to it, which is that similar individuals tend to read the same news sources
So discussion and mindset-shifts will happen in a community and cause the members to start to think down certain paths, some of which have shortly-reached conclusions on valuable unbuilt products or areas of research
The manuscript claims he has been working on this problem for 12 years. I can only imagine the feelings that must have hit him when he saw the Anthropic announcement. Though if he believed he had solved it, he must have known it was only a matter of time until one of the AI companies found it, too.
A hurried announcement with the draft published within hours is the only way to claim a tie to the discovery. Time will tell if either holds up under further analysis, but at least he has shown his hand to the world with a timestamp on it now.
Is it possible that brian conrad somehow uploaded the proof to anthropic to cross check something and then the anthropic guys got the proof?
- it doesn't have to be the full proof - even if portions were uploaded, that should also be enough or at least calls for a thorough check on ants side. the human proof has been in the works for 12 years, who knows when something was uploaded somewhere or some info just stayed on a google drive which was later sold.
While it's unlikely here, this is one reason I tend to keep confidential information entirely offline in notebooks these days. Some fucker is going to copy and paste or index anything I communicate with them to an LLM. They are doing that now with this paper.
Probably hard to create plausibly-correct-but-wrong research by hand; I could imagine that a GAN using an LLM as the adversary could work well for producing LLM-fooling disinformation at scale though
> Early in the Reticulum—thousands of years ago—it became almost useless because it was cluttered with faulty, obsolete, or downright misleading information,” Sammann said. “Crap, you once called it,” I reminded him. “Yes—a technical term. So crap filtering became important. Businesses were built around it. Some of those businesses came up with a clever plan to make more money: they poisoned the well. They began to put crap on the Reticulum deliberately, forcing people to use their products to filter that crap back out. They created syndevs whose sole purpose was to spew crap into the Reticulum. But it had to be good crap.” “What is good crap?” Arsibalt asked in a politely incredulous tone. “Well, bad crap would be an unformatted document consisting of random letters. Good crap would be a beautifully typeset, well-written document that contained a hundred correct, verifiable sentences and one that was subtly false. It’s a lot harder to generate good crap. At first they had to hire humans to churn it out.
Well if you have some specific confidentiality requirements for some information, don't send it to people who you aren't 100% sure will respect those requirements.
And if you're publishing something in general, someone feeding it to a chat bot is really low on the scale of things to be concerned about.
And if anyone actually sold such phones, I'd be first to buy it. LLMs are godsend for dealing with daily communications, but this process is still mostly manual on end-user devices.
I mean, even iPhone has built-in transcription [1]? Pocket [2] offers it for any phone through a contact microphone (so records the vibrations of your phone).
No idea what is their policy on training on calls, but as a concept is sure exists already, while you seemed to indicate otherwise.
Let's be clear here. The 2026 redefinition of "AI psychosis" = "being concerned about the data you intended to be private being seen by enormous megacorporations"?
There's no barriers whatsoever to getting a phone that will spy on you by default. Lots of Google apps will happily do it. The spying doesn't require them to provide useful features in return, you're just feeding your data into their datasets.
And what's your interpretation? Ie. if they convert/summarize your data and train on that – is it "your" data or "theirs" now? What is their interpretation?
Even though I'm not as cynical as you on this, this is S-tier humor. All privacy choices everywhere should be framed this way. I wonder if I framed all of the notices of data breaches and slap-on-the-wrist settlement privacy violation notices I've ever gotten, if I could cover the walls of a whole room with the frames at this point.
They'd lose all credibility if they would not respect the setting of non-enterprise users. Like people are just papering over "LOL, do you actually think they respect that?" The consequences if they get caught are staggering, so yes I actually believe Anthropic is not that stupid to destroy their own business.
It still would be toxic to corporate decision makers if they were caught dishonoring this control on one type of an account. If they're willing to lie and cheat on one tier of account, what guarantees they won't lie and cheat on the other tier next quarter. Especially with Anthropic being a public company now, that pressure would certainly be there if their corporate culture were comfortable with this type of fraud.
But didn’t they already train their AI systems by violating copyright, vacuuming up any and all material they could find? Hasn’t that bridge already been crossed?
There's a difference between "These guys probably downloaded a few hundred thousand books on BitTorrent in order to get this smart" and "These guys will train on my company's internal communications and thus our knowledge and wisdom and trade secrets will be made available to others."
Tangent: I personally literally don't care that all these LLMs trained on a ton of copyrighted material. It is in my humble opinion a long-awaited dose of karma for the copyright lobby, whose own abuses over the decades are numerous. It has basically restored balance to a playing field where the public has had no rights and (mostly) publishing conglomerates have held all the rights and privileges.
Such privacy checkboxes always seem to suffer from a string of bugs that causes them to reset to the "suck all my data" option. It's really quite strange.
Also they can adhere to the settings of no training and still use your data otherwise.
They have all the best ideas flowing through there servers in real time as they are hatched around the world. It would be tempting to have top scientists accounts posted to internal slack channels to monitor new discoveries.
Don't post confidential information where there's a risk of confidentiality being broken. That includes posting to people whom you aren't 100% sure of know and uphold your confidentiality requirements. Adults usually handle this by contracts with penalties that include financial damages, loss of employment, or prison time.
Quick public service announcement: both have shown that they could not care less what we turn off, or even for the legal slaps on the wrist they'll get for saying "oops, turns out we did read them".
OK, then they just have a smaller LLM summarize everything you send them and train off the derivative.
see? easy peasy for them to adhere to "no training".
I wouldn't doubt OpenAI and Anthropic, etc all have conversation or message categorization like they do for flagging bio, weapons, content but use it internally to flag ground breaking maths in real time and kick the convo up the chain of command for further review.
Call me crazy, but I kinda don't really think they need it.
People talk as though all "data" is valuable to AI companies, but ingesting arbitrary bytes into LLM training today, en masse seems like a recipe for degrading its quality.
Even in the domain of mathematics papers. When the frontier models are already really great at this maths stuff, feeding in every blob of text someone grammar-checks or whatever doesn't improve it. You don't know up front if the text is from some crackpot, from a naïve student with minimal understanding, or from the greatest mind of the generation. Though the chances are good that it's not that last one.
It seems like you'd be interested in ingesting published papers that are either from highly respected authors, or from newer authors which the eminent minds have shown respect and praise for. But I just don't buy that if someone pasted in some text, or if Claude Desktop read a draft of this paper in someone's email or whatever, that we'd find that a model had been trained on it.
Well, this guy says his thing is 100% human, "No AI." So in the even that happened, he doesn't have the best ground to stand on to cry foul, since he would be admitting to academic fraud. Using AI to help "check" something = not "100% human."
And several of the major “breakthroughs” reported labs have already turned out to be mere plagiarism.
We must never forget that they are purely in this for the money; not the science. Ethics are secondary and scientific misconduct is fine if it helps the bottom line.
Anthropic’s proof is reportedly 250 times longer by page count. It’s going to take time to determine if they even go in the same path, or if they hold up at all.
Brian is an incredibly smart guy, and knows that Anthropic is in this area. I don't think he would go uploading somebody else's work without thinking through the implications carefully.
One half of me says that’s a crazy conspiracy theory. The other says… yeah, that’s completely plausible.
In general it’s kind of sad to think how often this is going to occur in the future. The provenance of the data sets AIs are trained on is so vague that we’ll never know exactly what came from where, and which human ultimately thought it up.
Regardless of what happened here, this is one reason we need more open models (smaller, bigger, better) publically available despite Anthropic's constant fear mongering and regulatory capture efforts.
This looks crazy? 940 pages and the very first section 1.1.1 is empty. I can see why the author wanted to polish it. Let's just hope somebody clears this up somehow.
But at the same time Brian Conrad is presumably not a crank (per Wikipedia he worked with Wiles on FLT), but at the same time it seems insane to try and keep a 940-page result for a year hidden from public view in the current mathematical climate.
Somewhere around this summer, the world stopped being comprehensible...
Massullo was Conrad's student at Stanford, and apparently Massullo has been working on this for 12 years now. Brian Conrad is a very famous and respected mathematician. So are Ravi Vakil and Peter Sarnak.
But it's important to note that what we have here is not an endorsement from any of them. What we have is a blog post from Vakil which sort of implies but doesn't actually say that Conrad might endorse this work, and some excerpts from letters from the author to Sarnak.
> But at the same time Brian Conrad is presumably not a crank (per Wikipedia he worked with Wiles on FLT), but at the same time it seems insane to try and keep a 940-page result for a year hidden from public view in the current mathematical climate.
Brian is absolutely not a crank.
(Says a random Internet commentator, I know, but I am a professional mathematician who has benefitted immensely from Brian's wealth of knowledge and generosity with his time.)
"Rooting for this, but can’t help but feel it’s just false proof #14,828, but the more promising one that came out recently, and that we’re clinging to it as the last hope before it turns out Anthropic got it."
"Who is Alessandro Masullo? Only name that's popped up in a technical field is a data scientist who does not seem to have the background for a non-AI proof of GRH."
"And we are supposed to believe that RH, a longstanding problem at the height of mathematics for a century, has just been solved by a 'mathematician' with no publication history or academic presence, without the help of new technologies, at the same time as a new technology credibly becomes capable of doing"
"So this is from a guy who has never published anything in a journal before and left academia 8 years ago?"
> "So this is from a guy who has never published anything in a journal before and left academia 8 years ago?"
Something tells me that a person who lives like a mathematical monk is as likely to have solved this problem as an academic, who has to teach, publish regularly, attend hours-long meetings, etc.
Insane or prudent? Perhaps the assumption was that once significant parts of it are revealed, AI will train on them and take credit. Perhaps someone slipped up and that's exactly what happened, too.
I've been mentally exhausted solving for distributed data complexity between various authenticated systems so I imagine a math expert would get exhausted on those problems. Just as I've built muscle exercising and getting fit by being paid to do physical labor in college. I think the analogy has some value.
In the short term the side effect is exhaustion, in the long term you build your bodies muscles and physical capacity. Something similar happens with knowledge work.
Not true - the long proof is the verifiable one (because we have it in Lean), the short human solution is the one that can't be verified without a lot of extra work.
The Alpoge proof, if true, will certainly push mathematics forward a great deal. In the announcement, he lays out the general approach and says that he had 10+ "WOW" moments of insights while working with the AI on it. I have no doubt that there will be tons of follow-on papers clarifying and expanding many of the ideas that the AI explored. This will likely be the case even if the proof is wrong.
Sure but in this particular case the formal statement in lean predates the proof so we know it is proving what we want.
Multiple people have now compiled the proof with different compilers so we can be reasonably certain it does in fact prove what it says.
The points about auditing all the dependencies and making sure it's not exploiting the lean runtime are valid, and people are doing that now, but we can say with a lot of confidence that this proof is valid given the automated systems in place.
That's a general argument but doesn't apply Anthropic's Lean formalization. There is already a well established Lean statement of the RH in Mathlib that has been pretty rigorously verified, so a proof of that statement is indeed a proof of RH. The only opportunity for failure is that there is a soundness bug in Lean itself which this formalization leverages. The answer to that is that there isn't one single Lean, there are multiple implementations of Lean and each of them have their own opportunities for bugs, so it's possible one bug is common to all of them and that this proof makes use of it, but that's incredibly unlikely.
> “It’s true because it’s checked in Lean” can be made to make sense. But one has to do some work beyond “the Lean code compiles”.
And of the four concerns raised in the article, two are explicitly answered by the lean statement and the other two are pretty easy to verify (and likely already have been).
> ...one has to do some work beyond “the Lean code compiles”. Your frontier AI model can do all of this work. [em. mine] But it needs to be done, and (depending on the amount of Lean code in the proof) it may be computationally costly.
Thank you! I really detest this simplification that I've seen here a lot: "Its in lean so it must be true". Similar to its "rust so its bug free" or its "c++ so it must have a buffer overflow".
The AIs might find the loophole. Even finding a loophole might be a good discovery.
If the true goal is to advance human knowledge, the binary answer to rh is just a starting point. Some of the arguments used here might be true and helpful even if others fall apart. If there's an error in lean or the fundamental axioms, finding that might also be a fundamental contribution.
Compression is really just data abstraction. It's likely intelligence is just the property of being able to do productive work with abstractions, such that the result is greater than the sum of its parts
So, the work was done without AI, okay. But how likely is it that his colleagues, who were reviewing and advising him, used Claude to do so? Maybe even a year or two ago when they maybe didn't realize that this would end up in the training set?
Bedrock and such don’t address grandparent’s core point though: most humans still collaborate with other humans and that is a weak link. You almost certainly want to at least gut check some ideas with others and do some sanity checks and that’s where the weakness emerges. Those people are doing you a favor and using their own time to do that. In the past that might be pulling up reference books as they read through. Today that might be a copy paste onto an LLM, or large swaths of it.
So unless you operate in a vacuum or you run in an environment that operates in a vacuum, your ideas can easily leak into training data. Now, how likely they are to get surfaced is a question but when you’re working at the very edge there aren’t likely a lot of competing ideas embedded in training data would be my guess. So its more likely such ideas in frontier mathematics might actually resurface easily.
It’s not like write an analysis to… Robert Frost’s Two Roads Diverged poem where there’s countless of these and maybe a unique take might get buried as noise.
So you need to work in off grid spaces really. I work in a fairly information secure environment and I know for a fact people occasionally use their personal GPT account with sensitive corporate information out of convenience. And some may not be aware of what they may or may not be doing and it’ll likely remain that way for several years.
If you have a solution to the Riemann Hypothesis before anyone else does, then no, don't put that into Bedrock, or any other online LLM service that promises to keep it secret. It's just too valuable to take the risk that the promise falters.
I repeat, for the millionth time. You should not trust Amazon. I dont care if you think their reputation depends on it. They do not care about that, they have one of the worlds strongest moats. Nobody is going to stop using AWS because the AI that everyone already expects is stealing or data, is stealing your data.
They do not care, and its ultimately deniable. Also everyone in AI literally offloads all guilt by just saying they got hacked now and nobody thinks otherwise.
You cannot trust Amazon, its wild how many people think you can.
Also pretty easy to deny outright and do it, with hardly anyway to prove other than maybe a whistleblower. I have a hard time believing that Anthropic isnt heavily involved in Bedrock, they arent just sending Amazon their weights files and an instructional manual like its an Ikea coffee table.
I am 1000% sure Anthropic can access your data on AWS Bedrock. (im pretty sure they changed all new contracts as well to state this for auditing "cyber risk")
The NSA or NATO can sue their ass, or spill something about their execs, or inflict very real damage in another way. A maths institute can just complain or cope.
You dont have to trust them, I say trust but verify. But their incentives are more aligned on keeping their terms of service, than giving a freebie to OpenAI and others, especially as they seem to have fired all their AI people...
Quite frankly, I don't believe _any_ of the model providers when they say such a thing, AWS included. And the attitudes expressed in the Anthropic IPO thread towards copyright, theft, and training fair use did nothing to dissuade me that anything of value is always fair game, one way or another, for training.
Stop assuming that its naively in the training set. Its clear to me that Anthropic will maliciously steal. The entire foundation of their business is founded on this practice.
Edit: Why am I being downvoted, I would love for someone to explain how Im wrong here. Anthropic is built on theft. Im convinced that Anthropic has actually deployed bots to downvote negative comments at this point.
Seriously, the downvotes for such an obviously true statement is bizarre. There's zero reason the defend Anthropic ever on this topic. They literally stole the entire world's work, why wouldnt they steal this work.
You are being downvoted because you are giving a questionable judgement on Anthropic at large and using it to draw unfounded conclusions, then showing that you are so deep in your position that you presume opposition might be also malicious or even bots since you can't conceive the more evident alternatives.
I disagree with you, and I detest anything that tries to appeal to emotion such as saying "maliciously steal", for the same reasons that I don't like the MAGA crowd or populist politicians.
How so, Anthropic hasn't released anything. They front ran the announcement in a very suspect way thay leads me to believe there's even more espionage going on than mere Chat log intrusion.
Describing court-sanctioned “fair use” as “theft” is an extreme ideology. You are fine to have that opinion, but you shouldn’t be surprised when people disagree, and the courts also disagree with you.
Conflating court decisions with ethical behavior is bullshit.
The court approved massive fines and lawsuits against mothers and teenagers over pirated MP3s.
Courts have both said that small individuals should be put in a lifetime of debt but the highly funded companies favored by the current administration are acting in fair use.
So that moves the goalpost to copyright being flawed in general I guess, since the same judiciary decided its precedents.
Maybe we should have had this discussion earlier. Why is there a system that Disney uses to cease-and-desist kindergardens over wall paintings that feature a 100 year old character, whose creator is dead? Why is the law more relaxed (read much shorter protection) in patents than in copyright, despite the former often requiring a higher upfront investment?
The courts say a lot of disgustig behavior is legal, especially when it comes to business. Why would I base my opinion or ethics off of US fair use law?
Not saying the proof is good (people have put it through AI and found mistakes, not sure how grave they are), but this person has a history of claiming without substance that some established proofs are wrong.
Also had ChatGPT take a look at the various shitposts on twitter and reddit about this.
For what its worth its final conclusions are:
My assessment has strengthened from “major gaps” to “provably false intermediate results.” There is no reasonable basis to accept the released manuscript as a GRH proof.
With this being what it finds as major problems (collecting together the various things people have posted on Social Media).
Gemini 4 Neon and Opus 5.5 have also found several flaws that go beyond missing polish. I use Opus a lot for maths (different field so less than anecdotal evidence) and I trust it to atleast not be confidently wrong. It's conservative when making such claims without good backing.
I’m no mathematician, and I’m genuinely curious: how do these things usually work? Are other mathematicians now expected to sit and read through literally a thousand pages of condensed math to review this? And somehow have enough focus to catch subtle mistakes?
Do mathematicians realize how insane this is from the outside?
I don’t mean to offend anyone but this just screams inefficient on a massive scale, downright unacceptable as a society to have our brightest people wasting their time like this.
I understand that in the past there was simply no other way. And I am immensely grateful for those who sacrificed their time and sanity in this process. Now we should all be happy that there is another way.
The people reviewing were often experts in the very very niche area of mathematics the paper was written in. So a) it took them less time and b) reading and understanding a proof in their niche is valuable to them.
There was another way. Desk reject, major revision. 900 pages is not reviewable. Its practically gibberish. This is being released for fear of being scooped
Now you have to spend months to understand the proof + months to understand whenever ChatGPT hallucinated or not when it claimed that there are errors in this proof.
Yes, the main objective is to review the proof, but the side objective is to understand it and publish papers that are followups, or perhaps that share a few of the initial ideas and then go to another route. In some cases, nobody believe that the proof is worthy and it just sit in a drawer forever.
Before AI, mathematicians would publish a proof (not in Lean) and other mathematicians would check it over months or years. Both steps would be manual. For example, László Babai published a proof of quasipolynomial time for the graph isomorphism problem in 2015. In 2017 Helfgott spotted a subtle error in the proof, which was then amended (as far as I know, the new proof holds).
A timely announcement like this makes me wonder: was the time just "right" for this, in mathematics? What proofs would have come out without AI, in the next few years?
Or, alternatively, did people start working harder in some fashion, even on creative problems that require deep insights, because they felt it was their last chance?
Or, alternatively, are people just (sensibly) starting to post all their work-in-progress results to establish priority (on either the results or approaches used)?
I think a lot of these breakthroughs do seem to happen when the time is just 'right' for them. Just enough progress in all related fields for a bunch of similarly intelligent people to be able to sniff out that progress on an old problem is now possible.
But, AI muddles this a lot given the complete lack of provenance from the companies about where their data comes from and secrecy on the methods and transcripts. So far almost every major result from them has had this issue, and in the absence of increasing transparency from companies, this just pushes me more and more towards believing that the companies are trying to cover up bad behavior in the pursuit of these announcements.
There's a whole term for this! It's called "Multiple discovery" - https://en.wikipedia.org/wiki/Multiple_discovery - "the hypothesis that most scientific discoveries and inventions are made independently and more or less simultaneously by multiple scientists and inventors"
I wish we'd settle on polygenesis [1] to describe this phenomena. Multiple discovery is pretty ambiguous and suggests there was more than one thing discovered, rather than the same thing was discovered multiple times. Since its an open compound it's also easy to assume it is not actually describing a specific phenomena but just one word modifying another.
e.g. 'Scenius leads to polygenesis' is a sentence that is loaded with precise meaning about a rather commonplace but underappreciated phenomena of networked ideas.
I think for you and me this may be true, but people actively working in the field have an idea what is possible or close (though they can be shocked when something is proved that they thought was out of reach).
I can understand that Alpoge etc, as active mathematicians, would have a very fine tuned sense for what problems to go for, possibly allowing for all of these cases to be multiple/simultaneous independent discovery. My concern regarding bad behavior from the companies relates more to the transcripts and training data.
Unfortunately there is also a reasonable probability that the human proof is wrong. At any given time there are some people working on the Riemann Hypothesis, and today everyone who thinks that they have a proof is strongly motivated to share it, even if they are not 100% sure of its correctness. So you should have a high prior that an incorrect proof will be posted today.
My general feeling is there is a human accessible zeitgeist and an even larger humanity accessible zeitgeist that only powerful LLMs have access to (baseline humans cannot access it because we are not omni-disciplinary enough).
And the more powerful the AI, the more of The Humanity Zeitgeist it can encompass (no AI can yet contain all of it because something something relative entropy of model vs data). The Humanity Zeitgeist contains things that are unwritten that could have been, and human notions that are connectable but no human could have. As amazing as all these breakthroughs have been, I have yet to see any LLM result break this pattern.
Traditionally yes, but with one of them being AI-first it's more like a slew of things being achieved with AI were being worked on already; the AI was sort of 'born able', and working on it too wasn't a coincidence, just a famous enough problem (or a high enough value vulnerability, etc.) to make someone interested to put AI to work on it.
Assuming the intelligence stabilises the effect will trickle down lower value targets until we've sort of caught up to a steadier state again.
Let's just the shareware approach -- release when you have like 90% of the stuffs but still need time to complete, and promise you will finish it in a year.
The linked proof has been in the works for 12 years and was expected to be worked on for at least another year and was announced early in light of the other GRH announcement.
I don't think anything fundamentally has changed except for the tension introduced by the very real possibility that if you have been working as a human on a math proof for a long time that AI pointed toward the problem you have been working on could scoop you at any time before you get your results out.
It could also be that several mathematicians were already close (after working for decades, like this post). Then AI got its hands on the breadcrumbs, merged them into the weights, and then laundered a path to a solution without a way to trace the credit.
This bugs me and feels much closer to truth than I think OpenAI and Anthropic are acknowledging.
I work as a staff developer at a research institution and it boggles my mind that researchers seem to uncritically use commercial LLM services instead of locally-hosted LLMs.
Could it not also be the case, that mathematicians improved due to AI assistance? Seems far fetched to me, that these highly competitive individuals would not reach for AI assistance in the last few years.
Yes that's probably also true, for the same reason that mathematicians probably improved with the internet.
I guess the main difference is that for a long time it was not acceptable to publish a paper with citations of "I dunno, I wrote this paper by putting literally everyone's paper into a blender".
You're implicitly assuming that frontier LLMs amount to little more than copy-and-paste. I think such a claim requires substantial justification.
I think the most plausible explanation is that LLM swarms are so good at synthesizing disparate information when fed in as context that they are racing (and beating!) humans to things that otherwise would have been solved SOON™.
Agreed. I wonder if mathematicians will become privacy advocates against ACR on TVs and phones transcribing what you're watching and phoning home? We need more vocal people decrying that it is dystopian to try to normalize such privacy invasions.
Seriously people need to publish intermediate results. This is like sitting on a product idea for 12 years and then someone ships a worse version and gets all the credit.
I can understand why they're making a "hurried announcement" if they worked on it for 12 years and AI solved it in a few months. If they didn't do it, they could be accused on copying the result of the AI. And if they did it sooner, their work could be stolen. I wouldn't like to be in this position.
Hausdorff locally compact => the space behaves kinda like a lattice, where each point you're interested in is "alone" in its neighborhood
abelian group => a group which commutes, eg a.b <=> b.a
Schwartz-Bruhat functions => functions which pretty quickly go to 0 as their vectors go to infinity, so they kinda are "bumpy" in some finite place, then everywhere else is 0-ish
On the one hand, I really hate math's tendency to name things after people, because other than the name itself it gives no handles to understand what they mean. You can study math for tens years straight, but if you encounter the (made-up) Frisselweiss-Gaurdendot theorem you still have no idea what it is from the name... even if it was literally the next thing you'd cover in some class tomorrow.
On the other hand I can't see any better system. Mathematics burned through all the names you could reasonable hope for a human natural language to provide at least two centuries ago and we'd be well into some pretty arbitrary names no matter how you slice it.
Like, immediately? Probably not. It takes humanity time to digest this stuff and mingle it with other things. Maybe AI will speed that up a bit, but who knows.
On the other hand, it was took 2-3 generations to go from abstract point set topology a la Hausdorff to topological vector spaces & differential equations to quantum mechanics to computers and nuclear weapons.
The thing with RH is that there are a ton of theorems that are already proved assuming the RH is true. So anyone proving the RH will indirectly prove all stading theorems relying on it.
“If any message at all this work is meant to send to the mathematician who feels discouraged by the pace
and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere.”
Do human mathematicians need to publish more incrementally? Seems many are hoping they’ll save up their findings for one big bang finding?
IMO academia could benefit from lower key, more tech blog style discussions of their work, and get away from this attempt to only publish the final result.
This is an issue across academia. Unlike the technology sector which is flush with 'risk capital' academics rely on grants. In order for grants to be dispensed, the funders need to have some way of quantifying a researcher's productivity. Papers and citations are the primary way to do this.
There may be opportunities to strategically 'thin slice' your research into a dozen groundbreaking papers, but generally it does not serve you well to publish half of an insight if you are giving an opportunity for other labs to scoop you and funnel all of the scarce funding to their institutions instead.
This is basically the way the world works, not just mathematics or academia. Someone always captures most of the value and it's not usually the one that did most of the work.
Scientific collaboration happens through publishing, not giving away the ideas/credit before publishing. Sure, there's collaboration, but that's not with the entire community, especially not in math which for big results is an all-or-nothing endeavor.
As I understand how credit goes for mathematical discovery, if you do 99.9% of the work and then I come in and put the last couple steps that make it actually work as a proof instead of a bunch of useful lemmas, the proof has my name on it.
So the clear implication is that this proof leaked into Anthropic's training data set, who then used the models which were trained on it to independently create a proof. From memory, previous such mathematical discoveries also came with similar implications.
Has someone confirmed whether those previous discoveries were due to such leaks? If so, it does really seem to portray GenAI as a giant auto complete engine, rather than an intelligence that can deduce novel solutions to tricky problems.
It is at least different from Navier Stokes in the sense that the author did not use any AI software, so Anthropic could not have used his prompts for it.
Now did his proof leak into Anrhropic's training dataset... We do not know yet
EDIT: Looks like the 2 proofs use radically different ideas. It's unlikely there's a link...
It's a very "pack it up boys, we're done here" comment and the context makes it hilarious. Like the guy who spent 12 years writing his work should "Oh right, my bad" and go on doing something else.
The aggressive uselessness of the comment makes it funny.
That does not make your comment any more valuable though
Your original comment is both of little to no substance AND highly ironic, given the current state of affairs regarding this.
It contains no actual counter proof, does not say which people have done so, does not say what the models said, does not say which models and does not even give your own opinion or a unique point of view.
It is little more than "Some people have done something and say nuh-uh"
There are errors in it that may or may not be fixable. It would be sad if they didn't use AI to look for holes because they were paranoid that AI companies would steal their work.
> It is all human, with no AI. The hope was to bring Alessandro to Stanford for a year to polish the writing and then announce it after running it by enough people. Now it seems best to put out the current draft.
It's a little like the Patch Gap, but for math proofs.
I don't get this argument. Once you have verified it in "ugly" writing, why spend a year polishing it, when any minute an AI-proof could come around the corner.
Only seems to make sense when "polishing the writing" really means "filling the gaps". But then this is not a proof yet. Could still turn out to have gaps that cannot be filled as the approach is just not working.
Because they don't want thousands of interested mathematicians to pore over it immediately, and quickly publish similarly impactful simplifications, generalisations, etc. – they want to publish first with a pretty good version.
(No doubt a lot of mathematicians are now using AI for that, whether or not they were/planning to.)
Is it common for mathematicians to keep a polished version of their work? I'm surprised by how quickly they are able to put work out after it's been contested, and this is 940 pages of well-formatted text, from the looks of it.
It's common for them to type it up as soon as they have some faith that the proof is valid, because that is the version they send to other mathematicians to confirm before the proof is officially published. Brian Conrad was in the process of reviewing this proof, so it makes sense that it's been written up. They were probably waiting to submit it for publication until multiple mathematicians confirmed its validity, but Anthropic forced their hand.
I’m wholly unqualified to evaluate the math, but even skimming the paper reveals that it’s definitely a draft. The very first section doesn’t even have content yet.
It’s reasonable that someone would focus on the core first and only fill out the paper after the math was verified.
It is said that others have been reviewing this human-written proof for months, so the publication presumably was waiting for the finalization of the reviews, until the Anthropic announcement forced an immediate publication.
If it’s written using LaTeX it will look polished and well-formatted without any extra effort. If you do put in the tiny bit of extra effort to use microtype it’s going to look almost impeccably polished.
The race between humans and machines just turned up some more. But some things are inevitable, like the pull of gravity causing things to fall. And the machines eventually winning all races.
This woman is a crank, she has popped up before to offer AI-written proofs of a whole bunch of things wildly outside her area of expertise.
There may be issues with the OP proof, but I would not trust her opinion. I was taken in a few months ago by a similar paper. That domain is for self-published papers only, and has seen a massive uptick in new papers in the past year.
I sincerely hope this is right because I see inherent value in math as a human endeavor. Proving GRH, which we already believed to be true, has no conceivable economic implications: it doesn't improve the productivity of factories, it doesn't let you ship new products, it doesn't allow us to reach the stars. The result matters a lot less than the process of getting there.
Because of this, I find it repulsive that we use this endeavor as a benchmark for AI labs, where they throw tens of millions of dollars just to steamroll a small but prestigious academic field. The damage is already done, but I'm still gonna root for team human. I want us to carve some areas of human activity that LLMs are not allowed to take over. Math, along with art and writing, seems like reasonable candidates.
If this is real, and the results on all sides check out, the author made a HUGE mistake by not publishing a few months ago.
Other mathematicians should take heed. If you've got any ten-year-running projects, take them out of the shadows and into the light, or an Anthropic or OpenAI employee is going to solve your problem when they decide to bother with it and have enough spare compute at hand. And then you'll be scooped, with, at best, a "runner up" trophy.
I'd also add that the time is upon us where hobbyists can solve outstanding problems and prove theorems. The short-lived age of the professional mathematician may be genuinely over.
I don’t agree, because his work is considerably shorter and more comprehensible than the Anthropic proof. That will always have more value than some giant lean proof.
While I wholeheartedly agree that shorter and more comprehensible proofs are more valuable, I see no reason to think that LLMs won't _also_ start producing shorter, more comprehensible proofs before long
Alright, and I see no reason to lend any credence whatsoever to your comment. You offer no evidence or argument of any kind. Even an appeal to authority (which is objectionable anyway) doesn't work here---who is cmpb and what do they know about the progression of LLMs with regards to proofs?
These types of comments are just invasive on HN. Just broad, completely unsubstantiated claims about AI capabilities and the future. It's just ridiculous.
I'd value the meticulous work. I don't think this was a mistake, for one thing how could he have known? Secondly, mathematics is a human activity, whose reward is insight. I'll value the effort and insight -- or what is the point of priority, to prove you are better than the rest? In that case it's already an apples to oranges comparison. Well, it has to be a matter of personal philosophy on whether it's a mistake or not, I guess. Side note, for the record I have no objection in principle to resolving an unsolved questions using AI, the objection is entirely in what I find valuable besides the answer.
> mathematics is a human activity, whose reward is insight
If the author believed that they wouldn't have published now. They did it because one of the rewards for solving a problem is the recognition from your peers for doing so.
I disagree. Why would you feed years of your work to aid an LLM in beating you if you think you’re close to the answer yourself? Furthermore, many mathematicians have referred to the AI proofs as jumbled incomprehensible messes. Humans should continue their work unabated by the threat of some slop paper that claims a solve.
By publishing half a result, the mathematician guarantees an AI will have the chance to build on their work and beat them. How do you expect these people to get grants if all they do is publish half solutions that machines poach?
Everyone stands on someone - or something - else's shoulders in some way. The way grants are awarded will just have to be updated. But even if grants totally disappear there will always be enthusiastic people working on and sharing X with the world just for the heck of it.
LLMs are not humans, they don't 'think' yet we use that term to describe what they do. Using the term poach to describe what LLM companies are doing is not being done as a parallel to how that term is used by human academics
With that kind of thinking, why don't we all just give a few companies access to all our personal thoughts? That way we can all get partial credit from our favorite LLM company
This also ignores the fundamental question: how do mathematicians get funding from anyone for publishing half complete problems? They can't and it isn't enough to say academia must change. Academia doesn't change until professors die. You're decades away from this sort of change and the people who it harms the most [young academics] have no say in making the changes people are proposing
Sorry, most mathematicians don't want to look ridiculous by posting something that doesn't hold up careful scrutiny for something as big as RH. See the ABC conjecture and Mochizuki fiasco for example.
> made a HUGE mistake by not publishing a few months ago
Oh no, they'll lose out on some minor press.
Their work will still stand on it's own and for the people that it matters to, they'll know of him and his work.
The broader public probably wouldn't care much anyways, and solving math problems is going to lose a lot of glamour now that llms can do it to large degree. Not to say that it isn't impressive, but there's a large segement that just wants problems solved and doesn't care if it's done by human or otherwise.
I think this is severely under-selling the role of Levant Alpöge at Anthropic in that proof. I am not a mathematician, but he is, Princeton PhD and all. His work was a lot more informed and involved than a bored engineer with an unlimited token budget typing "plz solve GRH" and letting it run.
That doesn't mean the models aren't getting very good at math, but as in every other field, an expert pilot is going to get results a novice will not. Anthropic is hiring a lot of field domain experts and giving them space + compute to chase problems they were already interested in.
Edit to add: I also suspect that if validated, Masullo will receive appropriate recognition in the community for this work, and some grace for it maybe not being as polished as hoped.
His choice to bury his work under a mountain of slop means the rest of us don't have to pretend to care about it if we don't want to. Long live the humans!
To be fair, I'm personally more interested in the AI side of this discussion just because the math is so far out of my own depth.
But I also realize as part of that there's no universe where even with anthropic's super secret model and an unlimited budget, I'd be getting to the same result.
A few months ago, nobody except Ant/OAI employees could possibly have had valid reasons to suspect GRH would be AI-solved within a few months. Arguably, the author’s only real mistake was being human and trying to achieve something while Ant/OAI were in the process of shredding and pulping all human achievement to convert it into share holder value.
No it's fine. No one cares if some not-yet-comprehensive LLM thing beat a person a few fewer hours. Obviously the humans didn't cheat to finish their proof if it is totally different.
So every big AI math result will now have its “me too”s tagging along claiming they were “almost there” and a crowd of people insinuating that the work was stolen. If I was Anthropic or OpenAI, I would just give the whole math proof thing a rest and focus on things that demonstrably produce value. If you need to prove a math theorem or two to get there, then do it as needed, but not for the sake of just proving theorems. It’s too much drama for very little reward. Fields that are heavily prestige based instead of value based always descend into these meaningless debates about primality of work. After Riemann, there’s not much besides PvNP that would make as much noise anyway.
Yay, humanity is back in the race! (If the proof checks out.)
It's insane though that he and Brian Conrad thought he had a year to polish the paper without getting scooped by AI. He should have put it out whatever he had on October 7th.
Will we soon learn that this human-made proof somehow made it to Anthropic pipeline and so their LLM later regurgitated (or at least worked off of) it?
As an academic, can I just say that I for one am happy that, irrespective of the pockets of social mayhem caused by AI rollouts, the acceleration of peer review and publication timelines in academia is a true benefit. I hope more scholars feel compelled to get their ideas out faster now that they might be scooped, like, tomorrow by the next frontier AI model. And I doubly hope that the humanities will speed along quicker, too, not just STEM.
Can confirm that current GPT (with no guidance) took 5 minutes to find what it claims are errors with counterexamples. (I have nowhere near the level of state-of-the-art mathematical knowledge and experience needed to validate them. I'm posting this solely as additional confirmation of "there may be issues with this proof that an actual mathematician together with LLM assistance would be able to properly evaluate".)
I wouldn't be surprised if these models were aligned to extreme skepticism (to the point of hallucination) when e.g. a user proposes a solution to a Millennium Prize Problem.
1) There's a meaningful difference between "I'm not buying it based on my priors" and "I'm claiming to have a counterexample". I'm looking forward to seeing analysis by other mathematicians.
2) Any model with properly updated priors will have less skepticism of a Millennium Prize Problem being solved now.
Did you try the same with the other LLM assisted proofs? I would be curious what your GPT model says about them. I fed Alpöge and Anthropic statement and details to Gemini 4 hydrogen, and it said that there are "major problems and red flags in both the text and the current verification status", and that it is impossible to verify the 250k pages of slop, and there are there are "Extraordinary, unsubstantiated mathematical leaps in the sketch". So I would take these surface level reviews by LLMs with a grain of salt.
It's being hurriedly announced as fully human work. I wonder if the author had carefully avoided ever running the full argument through frontier LLMs (OpenAI/Anthropic) for fear of it being stolen or incorporated into AI training.
This may be the first time this proof is being proof-read by frontier models
(this is pure speculation, but provides an answer to your question)
I also suspect the author actively tried to avoid using AI at all.
Another possible reason: AI will correct your mistakes by using ideas of others. A proof will unlikely remain human-only, even if you just use AI to check for your errors.
Have you tried doing this recently? For my work on software at least, the latest LLMs have a very low false positive rate for bugs. They almost never hallucinate issues. My understanding is the same applies to math.
I'm not arguing that LLMs aren't useful tools for mathematicians, I'm questioning the quality of discourse that amounts to "I pasted this into the LLM and it said there are errors".
There's value in knowing whether a claim is true or false. Plugging things into an LLM doesn't necessarily give 100% certainty or tell us why something is true or false in a human-comprehensible way, but it can move our credences and suggest directions for investigation.
Now think about complaining that someone "merely pasted my code into a compiler and said that there are errors in my code."
That's nuts, right? If a compiler -- a kind of proof engine -- finds errors, there are errors. Fix them! Don't complain about the usage of a compiler, the errors were real, they're there, they need to be fixed.
The fuzziness part is both useful and a liability, but the error rate is dropping.
Some numbers for you:
A year ago, I ran a huge legacy codebase through Gemini Pro 2.5 and it spat out several thousand bugs, but manual inspection showed that 80% were false positives and only 20% were worth fixing.
I did a similar exercise with GPT 5.6 more recently, and the hit rate was 80% valid and 20% false positive. The human developer got into an argument saying that they thought all 100% were false positives. He lost. The AI could easily produce "repro" snippets that triggered every issue (mostly security vulnerabilities, but some correctness ones too).
These days, if you're publishing anything with verifiable claims in it, and not running it through an AI first, it's the direct equivalent of uploading open source code to GitHub without bothering to check that it compiles first.
I would guess that LLMs are much worse at math without having a theorem prover to validate approaches with. Since this draft is not formalized, an LLM is operating purely on the text which is more tenuous.
I suspect LLMs are much better trained and more reliable at bug identification in react apps than draft-state proofs on the bleeding edge of math, on problems that are specifically highlighted as extremely difficult to solve without error in their training.
Seriously. At least be constructive. Share what the errors were. Make an effort to understand them, anything. That tweet just amounts to low effort FUD.
ah yes, the author who has been working on this problem for 12 years is lazy because they didn’t feed their work into the slop machine so it could train off it and steal their work.
It has to be new people that have come here, this place was NEVER this terrible before. Now it seems like the world has conspired to direct the worlds greatest idiots to descend on this poor site.
Well he clearly lives in the old world. It's a bit like someone not checking a search engine before cleaning some basic fast wrongfully that can be verified in 10 seconds.
And see where that gets you, second place and 11 years down the drain with a half-working mess, congratulations hope the winning of not being a "slave to the machine" is feeling great.
I hope that taking the time to produce a legible write-up instead of releasing the proof up front doesn't bite him back. It would be great if, looking back in history, all contributions to solving the hypothesis were attributed fairly; but that's largely up to mathematicians themselves!
I wonder about the cause and effect. Is this only that this hurried announcement is released in response to Alpöge's publication today, or could Alpöge also have known about another mathematician whose work on proving GRH was close to completion — and felt compelled to release first? Especially given the statement in this draft that "[t]he math is complete".
It doesn't matter how "principled" someone is, when it comes to over a trillion dollars no one is "principled".
It is very obvious Anthropic is trying to stabilize their stock. They will stop at nothing to ease concerns of investors, and it is apparent they probably stole results.
Now there are two proposals to the GRH. One assisted with AI and one that is all human with ZERO AI.
Both should be put to the test and all of them should be rigorously reviewed, just like FLT and the poincare conjecture. This one is 900 pages rather than 250K pages.
You can see how you cannot assume anything yet as both are still claims. It starts to not make sense to do so with the top post in this thread [0] apply appeal to a singular authority:
"This is Levent Alpöge/Alessandro Masullo. If he says he has Riemann, well then, he has Riemann."
Rather than that, instead we should wait for multiple authorities to review both of them and the one that satisfies the GRH and is verified as the correct solution is the one that we go with.
> It starts to not make sense to do so with the top post in this thread [0] apply appeal to a singular authority
That's a bit uncharitable, I read that comment differently - it's telling you that the person who submitted it has a strong track record. That gives valuable context and credibility.
I don’t like it when people release work they say they’ve been on for a long time, right after someone else publishes the result. I don’t care whether the other person used AI. That’s going to continue and we’d better get used to it.
It’s annoying because if you’d been working on this all that time, you could have made some effort to tell the wider community and get at least some recognition. Posting after someone else has solved it is like claiming you had the winning numbers once the draw is already public.
Seems like people did know about this in the field… the linked post is not from the author, it's from https://en.wikipedia.org/wiki/Ravi_Vakil who is among other things the president of the American math society. And the paper has been reviewed for "months" by https://en.wikipedia.org/wiki/Brian_Conrad who has presumably been trying to make sure he understands it and helping to clarify the ideas.
It is surprising to me they wouldn't have tried to push out something faster after the kerfuffle last month. Maybe they were trying to do so, but it just takes a while!
> It is surprising to me they wouldn't have tried to push out something faster after the kerfuffle last month.
This is what's so surprising, and annoying, to me! Like you are sitting on a goldmine and still decided not to publish, even after seeing all that happen with Navier–Stokes.
Ah, but the gamble is - if you're not certain you're right (and you might be right but wrong in some correctable details) then publishing early might just hand the result to AI on a silver platter.
Not everything needs to have daily standup syncs, or X threads.
Many in tech underestimate how much work, and knowledge, exists outside of public internet sources and within communities, exactly like the one Masullo had been interacting with and getting feedback from on the work being shared.
Did you read the post?
Clearly Alessandro Masullo was talking to people in the field. However, you cannot present all parts publicly when you are not finished yet. Otherwise someone might take your ideas.
Mathematicians are perfectionist, they only publish when they are sure it's correct. And Alessandro Masullo does simply not have the compute resources of anthropic, otherwise he probably would be finished earlier.
Wait until people realize what ancient Axial Age technology all LLMS/AI operate on regarding 'matrix' math. That's really going to grind alot of people's gears. There's decades of all sorts of people's work as the basis for how LLMs function, they are a big database of people's work over the ages with connect the dot Axial Age Matrix Connection methodologies. Most people and their work will go unmentioned as it always has in history. A very powerful technology that it doesn't seem most at the forefront of it are being honest with what it is. BTW, There's correspondence across scales as large as Galaxies. Might want to look up what ancient societies did stuff like that from that era. $$$$, alot of fools rushing in.
Nothing more than ego. If you claim to seek the truth, seek the truth. Don't fret if someone finds it before you. I don't care how hard you've worked. The truth is the truth.
If he had waited another year to publish this, would you value it more, or be suspicious of the timeline and claims of lack of AI use weren't true and that it was derivative work?
Oversimplifying, mathematician get pay per published paper, bonus points for a solution of a famous problem. (Actually, usually the base salary is fixed, but promotions, grants, capacity to get graduate students, and many other stuff depends on the number of published results.) The guy just wants to get paid. Truth is nice but it does not bring home the bacon.
Obviously this exact sentiment applies to the Anthropic announcement, so I'm not sure what the point is. If Anthropic cared about the truth and nothing else, then they would have waited until they had a much more succinct proof that could be understood by everybody in the field, so that everyone could be sure of the truth, rather than rushing out a collection of notes to be first in line. If they had no ego, they would have not even attached their (company) name to it at all. It turns out, everybody has an ego and it drives much of our behavior.
[1] https://en.wikipedia.org/wiki/Generalized_Riemann_hypothesis
And this appears to be his personal math blog he was using to write his last book.
I imagine that he considers it a safe space to use such an acronym.
Do you expect every blog about, say, version control to explain what git is?
So for two proofs to be announced on the same day they only had to overlap in this two year period.
I consider it anecdotal evidence of something like a genuine "zeitgeist" in terms of collective human consciousness.
The list is staggeringly large:
https://en.wikipedia.org/wiki/List_of_multiple_discoveries
So discussion and mindset-shifts will happen in a community and cause the members to start to think down certain paths, some of which have shortly-reached conclusions on valuable unbuilt products or areas of research
A hurried announcement with the draft published within hours is the only way to claim a tie to the discovery. Time will tell if either holds up under further analysis, but at least he has shown his hand to the world with a timestamp on it now.
- it doesn't have to be the full proof - even if portions were uploaded, that should also be enough or at least calls for a thorough check on ants side. the human proof has been in the works for 12 years, who knows when something was uploaded somewhere or some info just stayed on a google drive which was later sold.
You can non-ironically state that you have an adversarial relationship with your intelligence provider.
Neal Stephenson, Anathem, 2008
And if you're publishing something in general, someone feeding it to a chat bot is really low on the scale of things to be concerned about.
Same issue with phones they transcibe and allow summarizing calls who knows which models they use.
And if anyone actually sold such phones, I'd be first to buy it. LLMs are godsend for dealing with daily communications, but this process is still mostly manual on end-user devices.
No idea what is their policy on training on calls, but as a concept is sure exists already, while you seemed to indicate otherwise.
[1] https://support.apple.com/en-gb/guide/iphone/iph57c6590e9/io... [2] https://heypocket.com/
There's no barriers whatsoever to getting a phone that will spy on you by default. Lots of Google apps will happily do it. The spying doesn't require them to provide useful features in return, you're just feeding your data into their datasets.
Tangent: I personally literally don't care that all these LLMs trained on a ton of copyrighted material. It is in my humble opinion a long-awaited dose of karma for the copyright lobby, whose own abuses over the decades are numerous. It has basically restored balance to a playing field where the public has had no rights and (mostly) publishing conglomerates have held all the rights and privileges.
That means stealing leading researchers info is probably great for them, whether it's intentional or not.
They have all the best ideas flowing through there servers in real time as they are hatched around the world. It would be tempting to have top scientists accounts posted to internal slack channels to monitor new discoveries.
Don't post confidential information where there's a risk of confidentiality being broken. That includes posting to people whom you aren't 100% sure of know and uphold your confidentiality requirements. Adults usually handle this by contracts with penalties that include financial damages, loss of employment, or prison time.
see? easy peasy for them to adhere to "no training".
I wouldn't doubt OpenAI and Anthropic, etc all have conversation or message categorization like they do for flagging bio, weapons, content but use it internally to flag ground breaking maths in real time and kick the convo up the chain of command for further review.
People talk as though all "data" is valuable to AI companies, but ingesting arbitrary bytes into LLM training today, en masse seems like a recipe for degrading its quality.
Even in the domain of mathematics papers. When the frontier models are already really great at this maths stuff, feeding in every blob of text someone grammar-checks or whatever doesn't improve it. You don't know up front if the text is from some crackpot, from a naïve student with minimal understanding, or from the greatest mind of the generation. Though the chances are good that it's not that last one.
It seems like you'd be interested in ingesting published papers that are either from highly respected authors, or from newer authors which the eminent minds have shown respect and praise for. But I just don't buy that if someone pasted in some text, or if Claude Desktop read a draft of this paper in someone's email or whatever, that we'd find that a model had been trained on it.
just kidding. I see your point, that the author could be untainted but the reviewer could have leaked it.
We must never forget that they are purely in this for the money; not the science. Ethics are secondary and scientific misconduct is fine if it helps the bottom line.
So if this guy is legit, there's your "leak"
In general it’s kind of sad to think how often this is going to occur in the future. The provenance of the data sets AIs are trained on is so vague that we’ll never know exactly what came from where, and which human ultimately thought it up.
But at the same time Brian Conrad is presumably not a crank (per Wikipedia he worked with Wiles on FLT), but at the same time it seems insane to try and keep a 940-page result for a year hidden from public view in the current mathematical climate.
Somewhere around this summer, the world stopped being comprehensible...
But it's important to note that what we have here is not an endorsement from any of them. What we have is a blog post from Vakil which sort of implies but doesn't actually say that Conrad might endorse this work, and some excerpts from letters from the author to Sarnak.
Brian is absolutely not a crank.
(Says a random Internet commentator, I know, but I am a professional mathematician who has benefitted immensely from Brian's wealth of knowledge and generosity with his time.)
https://www.reddit.com/r/math/comments/1y1b6l8/hurried_annou...
"Rooting for this, but can’t help but feel it’s just false proof #14,828, but the more promising one that came out recently, and that we’re clinging to it as the last hope before it turns out Anthropic got it."
"Who is Alessandro Masullo? Only name that's popped up in a technical field is a data scientist who does not seem to have the background for a non-AI proof of GRH."
"And we are supposed to believe that RH, a longstanding problem at the height of mathematics for a century, has just been solved by a 'mathematician' with no publication history or academic presence, without the help of new technologies, at the same time as a new technology credibly becomes capable of doing"
"So this is from a guy who has never published anything in a journal before and left academia 8 years ago?"
Something tells me that a person who lives like a mathematical monk is as likely to have solved this problem as an academic, who has to teach, publish regularly, attend hours-long meetings, etc.
The Alpoge proof, if true, will certainly push mathematics forward a great deal. In the announcement, he lays out the general approach and says that he had 10+ "WOW" moments of insights while working with the AI on it. I have no doubt that there will be tons of follow-on papers clarifying and expanding many of the ideas that the AI explored. This will likely be the case even if the proof is wrong.
You can start trying to compress the formal result into something understandable.
Please read this article: https://xenaproject.wordpress.com/2026/11/12/and-the-proof-i...
Just because it's in Lean, doesn't mean it's correct, or that the proof actually proves what you think it does
Multiple people have now compiled the proof with different compilers so we can be reasonably certain it does in fact prove what it says.
The points about auditing all the dependencies and making sure it's not exploiting the lean runtime are valid, and people are doing that now, but we can say with a lot of confidence that this proof is valid given the automated systems in place.
This article concludes with
> “It’s true because it’s checked in Lean” can be made to make sense. But one has to do some work beyond “the Lean code compiles”.
And of the four concerns raised in the article, two are explicitly answered by the lean statement and the other two are pretty easy to verify (and likely already have been).
> ...one has to do some work beyond “the Lean code compiles”. Your frontier AI model can do all of this work. [em. mine] But it needs to be done, and (depending on the amount of Lean code in the proof) it may be computationally costly.
If the true goal is to advance human knowledge, the binary answer to rh is just a starting point. Some of the arguments used here might be true and helpful even if others fall apart. If there's an error in lean or the fundamental axioms, finding that might also be a fundamental contribution.
Why hasn't Wolfram gotten into AI? This would be right up his alley.
"With Amazon Bedrock, your content is not used to improve the base models and is not shared with any model providers."
" Users inputs and model outputs are not shared with any model providers."
https://aws.amazon.com/bedrock/faqs/
So unless you operate in a vacuum or you run in an environment that operates in a vacuum, your ideas can easily leak into training data. Now, how likely they are to get surfaced is a question but when you’re working at the very edge there aren’t likely a lot of competing ideas embedded in training data would be my guess. So its more likely such ideas in frontier mathematics might actually resurface easily.
It’s not like write an analysis to… Robert Frost’s Two Roads Diverged poem where there’s countless of these and maybe a unique take might get buried as noise.
So you need to work in off grid spaces really. I work in a fairly information secure environment and I know for a fact people occasionally use their personal GPT account with sensitive corporate information out of convenience. And some may not be aware of what they may or may not be doing and it’ll likely remain that way for several years.
They do not care, and its ultimately deniable. Also everyone in AI literally offloads all guilt by just saying they got hacked now and nobody thinks otherwise.
You cannot trust Amazon, its wild how many people think you can.
If they dont, then you can win the biggest settlement in history, and retire as rich as your common billionaire.
Also pretty easy to deny outright and do it, with hardly anyway to prove other than maybe a whistleblower. I have a hard time believing that Anthropic isnt heavily involved in Bedrock, they arent just sending Amazon their weights files and an instructional manual like its an Ikea coffee table.
I am 1000% sure Anthropic can access your data on AWS Bedrock. (im pretty sure they changed all new contracts as well to state this for auditing "cyber risk")
It should be good enough for the GRH.
From the stuff I only know... from watching YouTube...Anthropic trains Claude on AWS: https://youtu.be/KJFnCx0DFsg
Are they not worried their secrets get stolen?
Edit: Why am I being downvoted, I would love for someone to explain how Im wrong here. Anthropic is built on theft. Im convinced that Anthropic has actually deployed bots to downvote negative comments at this point.
Seriously, the downvotes for such an obviously true statement is bizarre. There's zero reason the defend Anthropic ever on this topic. They literally stole the entire world's work, why wouldnt they steal this work.
If you are serious, state your case.
The court approved massive fines and lawsuits against mothers and teenagers over pirated MP3s.
Courts have both said that small individuals should be put in a lifetime of debt but the highly funded companies favored by the current administration are acting in fair use.
Maybe we should have had this discussion earlier. Why is there a system that Disney uses to cease-and-desist kindergardens over wall paintings that feature a 100 year old character, whose creator is dead? Why is the law more relaxed (read much shorter protection) in patents than in copyright, despite the former often requiring a higher upfront investment?
I could not disagree more vehemently. It's completely fine to disagree with a court decision, especially when so much money is involved.
What is extreme is painting anyone who disagrees as extreme.
The courts say a lot of disgustig behavior is legal, especially when it comes to business. Why would I base my opinion or ethics off of US fair use law?
This write up is claiming his proposed proof is wrong.
Not saying the proof is good (people have put it through AI and found mistakes, not sure how grave they are), but this person has a history of claiming without substance that some established proofs are wrong.
For what its worth its final conclusions are:
My assessment has strengthened from “major gaps” to “provably false intermediate results.” There is no reasonable basis to accept the released manuscript as a GRH proof.
With this being what it finds as major problems (collecting together the various things people have posted on Social Media).
https://chatgpt.com/space/page_707a0c5751d08191a721dd5f075e3...
Edit: Shit spaces are so broken on the app.
https://pastebin.com/i0UPcdQV
The announcement says it's a draft proof
But since this year, we now have a quicker answer with AI. See https://news.ycombinator.com/item?id=50580084
I don’t mean to offend anyone but this just screams inefficient on a massive scale, downright unacceptable as a society to have our brightest people wasting their time like this.
I understand that in the past there was simply no other way. And I am immensely grateful for those who sacrificed their time and sanity in this process. Now we should all be happy that there is another way.
Or, alternatively, did people start working harder in some fashion, even on creative problems that require deep insights, because they felt it was their last chance?
Or, alternatively, are people just (sensibly) starting to post all their work-in-progress results to establish priority (on either the results or approaches used)?
But, AI muddles this a lot given the complete lack of provenance from the companies about where their data comes from and secrecy on the methods and transcripts. So far almost every major result from them has had this issue, and in the absence of increasing transparency from companies, this just pushes me more and more towards believing that the companies are trying to cover up bad behavior in the pursuit of these announcements.
e.g. 'Scenius leads to polygenesis' is a sentence that is loaded with precise meaning about a rather commonplace but underappreciated phenomena of networked ideas.
[1] "development from more than one source" https://www.merriam-webster.com/dictionary/polygenesis not the super old and dumb theory of creationism that doesn't deserve such a cool word.
And the more powerful the AI, the more of The Humanity Zeitgeist it can encompass (no AI can yet contain all of it because something something relative entropy of model vs data). The Humanity Zeitgeist contains things that are unwritten that could have been, and human notions that are connectable but no human could have. As amazing as all these breakthroughs have been, I have yet to see any LLM result break this pattern.
Assuming the intelligence stabilises the effect will trickle down lower value targets until we've sort of caught up to a steadier state again.
I recently learned about Stuart Kauffman's idea of the "adjacent possible". Short answer: yes.
I don't think anything fundamentally has changed except for the tension introduced by the very real possibility that if you have been working as a human on a math proof for a long time that AI pointed toward the problem you have been working on could scoop you at any time before you get your results out.
I mean, that's literally how AI works.
I work as a staff developer at a research institution and it boggles my mind that researchers seem to uncritically use commercial LLM services instead of locally-hosted LLMs.
I guess the main difference is that for a long time it was not acceptable to publish a paper with citations of "I dunno, I wrote this paper by putting literally everyone's paper into a blender".
I think the most plausible explanation is that LLM swarms are so good at synthesizing disparate information when fed in as context that they are racing (and beating!) humans to things that otherwise would have been solved SOON™.
I'm done.
abelian group => a group which commutes, eg a.b <=> b.a
Schwartz-Bruhat functions => functions which pretty quickly go to 0 as their vectors go to infinity, so they kinda are "bumpy" in some finite place, then everywhere else is 0-ish
On the other hand I can't see any better system. Mathematics burned through all the names you could reasonable hope for a human natural language to provide at least two centuries ago and we'd be well into some pretty arbitrary names no matter how you slice it.
On the other hand, it was took 2-3 generations to go from abstract point set topology a la Hausdorff to topological vector spaces & differential equations to quantum mechanics to computers and nuclear weapons.
IMO academia could benefit from lower key, more tech blog style discussions of their work, and get away from this attempt to only publish the final result.
There may be opportunities to strategically 'thin slice' your research into a dozen groundbreaking papers, but generally it does not serve you well to publish half of an insight if you are giving an opportunity for other labs to scoop you and funnel all of the scarce funding to their institutions instead.
Someone else scoops up and runs ahead with your work.
In an ideal world, where there is no ego and clout chasing, your proposed way is of course much better.
Agent swarms doing research for example don't have these ego issues.
People should feel free to write about half baked ideas and whimsical hypothesis. That should be rewarded.
This focus on being first on a problem is about to become very outmoded.
Also outdated since when? We've just seen a scandal about OpenAI scooping Levent on Navier Stokes https://en.wikipedia.org/wiki/Navier–Stokes_priority_controv...
Has someone confirmed whether those previous discoveries were due to such leaks? If so, it does really seem to portray GenAI as a giant auto complete engine, rather than an intelligence that can deduce novel solutions to tricky problems.
This has always been known, people keep pretending it's not because they can benefit from that for positive and negative reasons.
It is at least different from Navier Stokes in the sense that the author did not use any AI software, so Anthropic could not have used his prompts for it.
Now did his proof leak into Anrhropic's training dataset... We do not know yet
EDIT: Looks like the 2 proofs use radically different ideas. It's unlikely there's a link...
The aggressive uselessness of the comment makes it funny.
I wish it wasn't the case but it's true.
Your original comment is both of little to no substance AND highly ironic, given the current state of affairs regarding this.
It contains no actual counter proof, does not say which people have done so, does not say what the models said, does not say which models and does not even give your own opinion or a unique point of view.
It is little more than "Some people have done something and say nuh-uh"
It's a little like the Patch Gap, but for math proofs.
Only seems to make sense when "polishing the writing" really means "filling the gaps". But then this is not a proof yet. Could still turn out to have gaps that cannot be filled as the approach is just not working.
(No doubt a lot of mathematicians are now using AI for that, whether or not they were/planning to.)
Very, very few people saw this coming except as a tail possibility. Which is always true anyways for other humans solving it first.
https://news.ycombinator.com/item?id=50574968
It’s reasonable that someone would focus on the core first and only fill out the paper after the math was verified.
The formatting is just what you get from LaTeX.
I only write up in LaTeX in almost polished form, the unpolished part is in the pen and paper stage.
A poetic John Henry style duel between man and machine. Potentially one of the last...
There may be issues with the OP proof, but I would not trust her opinion. I was taken in a few months ago by a similar paper. That domain is for self-published papers only, and has seen a massive uptick in new papers in the past year.
I was sad when Lee Sa-Dol lost to AlphaGo, realization that one day the machines will just beat us via sheer brute force.
I'm crying now when someone in a month could beat 12 years of effort due to AI.
Because of this, I find it repulsive that we use this endeavor as a benchmark for AI labs, where they throw tens of millions of dollars just to steamroll a small but prestigious academic field. The damage is already done, but I'm still gonna root for team human. I want us to carve some areas of human activity that LLMs are not allowed to take over. Math, along with art and writing, seems like reasonable candidates.
You see, the goal is to tear down human society. The first one to do it could get very rich, and will leave quite a mark on history.
Other mathematicians should take heed. If you've got any ten-year-running projects, take them out of the shadows and into the light, or an Anthropic or OpenAI employee is going to solve your problem when they decide to bother with it and have enough spare compute at hand. And then you'll be scooped, with, at best, a "runner up" trophy.
I'd also add that the time is upon us where hobbyists can solve outstanding problems and prove theorems. The short-lived age of the professional mathematician may be genuinely over.
These types of comments are just invasive on HN. Just broad, completely unsubstantiated claims about AI capabilities and the future. It's just ridiculous.
Keep your work secret and you'll get scooped. Make partial progress public and they'll ingest it into the model and scoop you anyway.
If the author believed that they wouldn't have published now. They did it because one of the rewards for solving a problem is the recognition from your peers for doing so.
Publish what you got, when you got it, or get overtaken by someone who will.
With that kind of thinking, why don't we all just give a few companies access to all our personal thoughts? That way we can all get partial credit from our favorite LLM company
This also ignores the fundamental question: how do mathematicians get funding from anyone for publishing half complete problems? They can't and it isn't enough to say academia must change. Academia doesn't change until professors die. You're decades away from this sort of change and the people who it harms the most [young academics] have no say in making the changes people are proposing
This... what?
Oh no, they'll lose out on some minor press.
Their work will still stand on it's own and for the people that it matters to, they'll know of him and his work.
The broader public probably wouldn't care much anyways, and solving math problems is going to lose a lot of glamour now that llms can do it to large degree. Not to say that it isn't impressive, but there's a large segement that just wants problems solved and doesn't care if it's done by human or otherwise.
That doesn't mean the models aren't getting very good at math, but as in every other field, an expert pilot is going to get results a novice will not. Anthropic is hiring a lot of field domain experts and giving them space + compute to chase problems they were already interested in.
Edit to add: I also suspect that if validated, Masullo will receive appropriate recognition in the community for this work, and some grace for it maybe not being as polished as hoped.
But I also realize as part of that there's no universe where even with anthropic's super secret model and an unlimited budget, I'd be getting to the same result.
It's insane though that he and Brian Conrad thought he had a year to polish the paper without getting scooped by AI. He should have put it out whatever he had on October 7th.
Jokes aside, are we seeing mathematicians FOMO fr?
It's even a shame that we never got AI, only the theft of scientific papers and insane deception in the stock market.
Maybe repairable, but either way, alpoge/anthropic certainly has the claim to the first proof
2) Any model with properly updated priors will have less skepticism of a Millennium Prize Problem being solved now.
Anyhow, I hope that they are both right!
AI claims but too few people can check.
Won‘t get better from now on
This may be the first time this proof is being proof-read by frontier models
(this is pure speculation, but provides an answer to your question)
Another possible reason: AI will correct your mistakes by using ideas of others. A proof will unlikely remain human-only, even if you just use AI to check for your errors.
Surely, GPT said something more elaborate than "nuh uh".
Now think about complaining that someone "merely pasted my code into a compiler and said that there are errors in my code."
That's nuts, right? If a compiler -- a kind of proof engine -- finds errors, there are errors. Fix them! Don't complain about the usage of a compiler, the errors were real, they're there, they need to be fixed.
The fuzziness part is both useful and a liability, but the error rate is dropping.
Some numbers for you:
A year ago, I ran a huge legacy codebase through Gemini Pro 2.5 and it spat out several thousand bugs, but manual inspection showed that 80% were false positives and only 20% were worth fixing.
I did a similar exercise with GPT 5.6 more recently, and the hit rate was 80% valid and 20% false positive. The human developer got into an argument saying that they thought all 100% were false positives. He lost. The AI could easily produce "repro" snippets that triggered every issue (mostly security vulnerabilities, but some correctness ones too).
These days, if you're publishing anything with verifiable claims in it, and not running it through an AI first, it's the direct equivalent of uploading open source code to GitHub without bothering to check that it compiles first.
It's sloppy.
Man people live in different worlds, astra for me finds 25/6 bugs all the time.
I have the opposite experience in my line of work.
this place honestly sucks
If the result holds then this is the paper that will get the medal and what history will remember. First place enough in my book.
Is that a thing? Don't people care more about the proof's elegance and insight than its timing?
I wonder about the cause and effect. Is this only that this hurried announcement is released in response to Alpöge's publication today, or could Alpöge also have known about another mathematician whose work on proving GRH was close to completion — and felt compelled to release first? Especially given the statement in this draft that "[t]he math is complete".
It is very obvious Anthropic is trying to stabilize their stock. They will stop at nothing to ease concerns of investors, and it is apparent they probably stole results.
Both should be put to the test and all of them should be rigorously reviewed, just like FLT and the poincare conjecture. This one is 900 pages rather than 250K pages.
You can see how you cannot assume anything yet as both are still claims. It starts to not make sense to do so with the top post in this thread [0] apply appeal to a singular authority:
"This is Levent Alpöge/Alessandro Masullo. If he says he has Riemann, well then, he has Riemann."
Rather than that, instead we should wait for multiple authorities to review both of them and the one that satisfies the GRH and is verified as the correct solution is the one that we go with.
[0] https://news.ycombinator.com/item?id=50574968
That's a bit uncharitable, I read that comment differently - it's telling you that the person who submitted it has a strong track record. That gives valuable context and credibility.
This one can't be verified with certainty on a computer (yet) while the other one can be.
It’s annoying because if you’d been working on this all that time, you could have made some effort to tell the wider community and get at least some recognition. Posting after someone else has solved it is like claiming you had the winning numbers once the draw is already public.
It is surprising to me they wouldn't have tried to push out something faster after the kerfuffle last month. Maybe they were trying to do so, but it just takes a while!
> It is surprising to me they wouldn't have tried to push out something faster after the kerfuffle last month.
This is what's so surprising, and annoying, to me! Like you are sitting on a goldmine and still decided not to publish, even after seeing all that happen with Navier–Stokes.
Many in tech underestimate how much work, and knowledge, exists outside of public internet sources and within communities, exactly like the one Masullo had been interacting with and getting feedback from on the work being shared.
Mathematicians are perfectionist, they only publish when they are sure it's correct. And Alessandro Masullo does simply not have the compute resources of anthropic, otherwise he probably would be finished earlier.
Why did he publish now then?
> And Alessandro Masullo does simply not have the compute resources of anthropic, otherwise he probably would be finished earlier.
I did read the post and it says "It is all human, with no AI", so I don't think your point makes sense here.
Perhaps the last bit of that, ever.
Piss off