MIDREAL
Menu

Levent Alpöge claims proof of the Generalized Riemann Hypothesis [pdf]

https://alpo.ge/remark.pdf (alpo.ge)

2870 points | 42h ago | Discuss on Hacker News | Back to Radar

Comments

jrflo 42h ago on HN
Insane if true! Would also be great for Alpoge after being scooped by OpenAI on Navier Stokes...

Edit: just saw the Lean code is included in the footnote. Considering the author's reputation, I think it's fair to remove the qualifications from the post title, by the standards of today, it's proven.

hgoel 41h ago on HN
This may be how the court of public opinion often works, but plenty of otherwise very prolific mathematicians have claimed to solve problems only to turn out to have been mistaken.
jrflo 41h ago on HN
This is true, but since there is a lean formalization the odds of that are much lower. Only the correctness of the problem statement needs to be verified, which I would assume is fairly easy for RH? But I am not an expert, so would be curious to hear from someone who is.
hgoel 41h ago on HN
I'm not an expert in this area either, so I too will be waiting for experts to review things, but I will remain skeptical until then.

After the excitement of the Navier-Stokes solution, only for it to eventually seem to not really solve the full problem and leaving discourse about the problem in a very muddy state, I'd rather wait and see if it turns out this proof also makes some assumptions that ultimately limit its impact.

It is probably still some form of breakthrough, but I will expect it to not be as big as a full general proof of the Riemann Hypothesis, until proven otherwise.

Edit: A thought though, if this proof is correct and can't be made much shorter than the current ~250k pages, I wonder if the result has more value as a sign that human accessible proofs to these kinds of hypotheses were stuck behind some mathematical abstraction/tool that we haven't quite discovered yet? Sort of like how you can brute force algebra with real numbers for working with AC circuits, but it's pretty ugly and painful compared to using phasors. If so, maybe the real contribution from AI-driven high profile proofs would come from mathematicians studying the proofs and devising tools to make them more accessible.

azan_ 41h ago on HN
> After the excitement of the Navier-Stokes solution, only for it to eventually seem to not really solve the full problem

The version of solved problem is exactly the one that's permitted by Clay Institute. If it wasn't done by AI there would be no such discussion and everyone would just accept it as a huge breakthrough.

QuesnayJr 39h ago on HN
The Clay problem probably should have been "Either forced proof, or unforced counterexample", since those are the two most interesting versions.
inkysigma 40h ago on HN
Buckmaster is apparently claiming unforced NS as of today:

https://github.com/tristanbuckmaster/navier-stokes

yewenjie 42h ago on HN
> The argument is incredibly long at the moment, and while its formalization is finished and it is routine to produce a huge, unreadable document containing sufficient detail, in my view of my path I have just taken a step. I intend to do my best to understand and impart understanding of this argument and its certain simplifications, to speak about it, to write it properly.

Okay, I am anxious whether this holds true or not, but he claims that the formalization is finished.

Why not just release the formalization in the meantime?

This is probably the biggest day in math in recent memory.

Edit: nevermind, the Lean formalization is linked.

csheehan10 42h ago on HN
He has linked to the Lean formalization in this note.
jamessb 42h ago on HN
> Why not just release the formalization in the meantime?

I think this is what is linked to in footnote 1, inside the 14-part tar.gz archive.

asa123 42h ago on HN
doesnt he link to a lean file in the pdf?
scottmsul 41h ago on HN
He did, see the links below.

https://alpo.ge/lean.tar.gz.part01 ... https://alpo.ge/lean.tar.gz.part14

Though he decided against releasing a slop-filled pdf, which would have been over 250,000 pages.

programjames 41h ago on HN
Biggest day so far.
mellosouls 41h ago on HN
<pedantic>

Implied in "recent memory", which by definition includes everything up to the point of that comment, but not beyond.

</pedantic>

krona 41h ago on HN
> Why not just release the formalization in the meantime?

Nobody has a machine with 10TB of RAM required to run the Lean proof?

josefx 41h ago on HN
Throw some swap space at it?
cillian64 22h ago on HN
Apparently 50-100GB RAM required: https://math.mit.edu/~drew/grh/

A decent bit, but not totally crazy.

jmyeet 41h ago on HN
I'm old enough to remember when Andrew Wiles announced a proof of Fermat's Last Theorem way back in 1993 [1]. It was long and complex and took at least 2 years for the maths community to accept it.

I wouldn't be surprised if this isn't similar (or worse). The days of short, succinct, elegant, intuitive proofs is probably over.

[1]: https://en.wikipedia.org/wiki/Wiles%27s_proof_of_Fermat%27s_...

ilaksh 41h ago on HN
On wikipedia it says it's 250,000 pages. That might be an error on the page though.
electroweak 41h ago on HN
I wonder if that will be the last major result proved only by a human mind.
wiosnaintel 40h ago on HN
Poincaré conjecture was proven in 2002.
sebzim4500 39h ago on HN
For what it's worth, the lean proof is a similar length (15M lines vs 12M lines) to Anthropic's first formalisation of FLT so at least it probably isn't much more complicated.

God knows what will happen when we have billion line lean proofs that aren't easily shortened. They might just be forever outside human understanding.

kzrdude 35h ago on HN
Well, more than one year of that was Wiles and Taylor working to fix a problem in the proof.
rcpt 32h ago on HN
Unfortunately there's reason to believe that science as a whole has just become too hard in recent decades. Not as much low hanging fruit across all fields.
xianshou 42h ago on HN
This is Levent Alpöge (https://en.wikipedia.org/wiki/Levent_Alp%C3%B6ge), a supremely qualified mathematician and one of the best in the world at guiding AI towards new discoveries in mathematics.

If he says he has Riemann, well then, he has Riemann.

charles_f 42h ago on HN
Q.e.d.
solyin 42h ago on HN
Quod Erat Demonstrandum, meaning "that which was to be demonstrated" or "what was to be shown"
away0g 41h ago on HN
TBD
bitlax 41h ago on HN
Wow I just thought those were the initials of the guy who wrote every proof ever.
mockerell 41h ago on HN
Don’t forget his French cousin CQFD
ben_w 41h ago on HN
C'est quatre fromages dauphinoise.
charles_f 31h ago on HN
C qoi fo démontrer?
bee_rider 41h ago on HN
This would be a good Easter Egg for a story. Mathemystics who worship the ancients, Pythagoras and the mysterious QED.
mikrl 41h ago on HN
Characteristic of the disciples of QED was their renunciation of proper names in favour of a black rectangle.
igleria 41h ago on HN
We used to joke at university that it meant "queda entonces demostrado"
darkwater 41h ago on HN
I believe it's actually the correct "reverse acronym" in Spanish? I mean, keeping the Latin one they created a Spanish version that has the same meaning, in Spanish.

In Italian we use "CVD" for it (Come Volevasi Dimostrare)

thaumasiotes 40h ago on HN
CVD seems to be a fairly accurate translation of QED. The Latin is "what was supposed to be shown"¹ and it looks like the Italian means something like "as it was desired to show"? The Italian isn't a relative clause, but the meaning looks basically the same.

The Spanish appears to mean something more like "and now, it is shown", which is a completely different meaning expressed with unrelated words that happen to start with the same letters. (Spanish still has direct equivalents of all three Latin words², but I don't think que era demostrando even means anything.)

¹ In an English proof, you might see something like this:

"We can then see that a = b < c = d ≤ e, and therefore a < e, which is what we wanted to prove."

QED stands for the "what we wanted to prove" part of that, except that there is no we; in the Latin, we simply make reference to the fact that something was supposed to be proven.

² Sticklers might argue that Spanish has lost Latin's quod and replaced it with what is really a descendant of quid. They are almost the same word; quid is an interrogative form.

eriquesito 41h ago on HN
Isn't it what it literally means in spanish?
racl101 41h ago on HN
I think the joke is that while other cultures find Latin to sound esoteric and/or fancy, especially if they had to take courses to learn it, that your garden variety, non-educated, Spanish speaking person just simply gets it cause it's so damn similar to Latin that it doesn't faze them.

(Spanish speaker btw).

igleria 40h ago on HN
ding ding ding correct!
thaumasiotes 40h ago on HN
Really? The Spanish words don't seem to bear any relationship to the Latin words, so I fail to see how the relatedness of the languages could be helping.

Have you studied Latin, and if not, what does "quod erat demonstrandum" appear to mean to you?

dmurray 39h ago on HN
In English we have this with Rest In Peace.
thaumasiotes 35h ago on HN
That doesn't look similar. Rest in peace is almost a direct word-for-word translation of requiescat in pace "let [whoever] rest in peace".

By contrast, queda entonces demostrado is totally unrelated to quod erat demonstrandum, using different words (except that the third words are both forms of the same verb, very different forms) to express a different meaning.

tancop 39h ago on HN
Not exactly. The Spanish version translates to something like "so it is proved" and is a full sentence, but the Latin is a phrase closer to "which was supposed to be proved".
layer8 41h ago on HN
You don't say.
bell-cot 40h ago on HN
Quantum Electrodynamics sounds cooler.
ilaksh 41h ago on HN
He did post the Lean proof.

Also the Wikipedia says that the full written proof is 250,000 pages. Which I am guessing might be a mistake. If not it could take a while for humans to verify it.

well_ackshually 41h ago on HN
[flagged]
hackyhacky 41h ago on HN
> Even if their discovery happens to be wrong. (lol navier stokes)

Has it in fact been shown to be wrong?

well_ackshually 41h ago on HN
There are dozens of papers claiming that their formalization is wrong, that their natural language claims do not match the lean version of it, some parts of the proof are insufficient. finally, their entire proof relies on an external force. For a problem that is about, for 99% of people that care about it... unforced fluids.

but hey, it gets them a month of being talked about.

selectodude 40h ago on HN
tomhow 26h ago on HN
We've banned this account. You've had more than enough opportunity to demonstrate that you intend to use the site earnestly.
bonzini 41h ago on HN
He says it in a footnote, probably referring to ~2M lines of Lean?
CapnTrippy 40h ago on HN
:) 250k pages?! I could believe that, maybe. It might amount to -

"It's true, because the rest of maths is true"

Which I suppose, in any consistent universe, must be true for all proofs, just that up until now, our proofs have been relatively simple.

Interesting question occurred to me - how long can any proof get before it can be simplified?

I suspect arbitrarily long, even for problems that can be simply stated.

simonw 42h ago on HN
Levent was one of the two mathematicians working on Navier–Stokes when OpenAI published their work on that problem - which was produced in a rush when they heard a rumor that someone from Anthropic had promising results on a millenium prize problem.
QuesnayJr 41h ago on HN
It's a sign of the times that when someone with an AI beats you to one Millenium Problem, you can just turn the attention of your AI to another Millenium Problem.
finghin 41h ago on HN
It's almost certain that they have had people working on Riemann since way before the Navier-Stokes theorem proof was nearing completion. It's a long shot, but the value of a proof is next-level compared to NS
thaumasiotes 40h ago on HN
> It's a long shot, but the value of a proof is next-level compared to NS

The value of a proof? People have assumed it's true for so long that there are any number of results already proved "if the Riemann hypothesis is true".

What value does the proof add?

Jweb_Guru 40h ago on HN
The value it adds is that those theorems are now actually true, which means all those derived results are "real" mathematics... as every mathematician knows, it is quite easy to prove very surprising theorems when False is in your context.
thaumasiotes 39h ago on HN
But that's 100% wrong. If it was easy to prove surprising theorems from an assumption of the Riemann hypothesis, it would also be easy to disprove the Riemann hypothesis.
xrisk 39h ago on HN
not at all. we can prove a lot of things using P != NP, but that doesn’t make a proof or disproof of that statement easy
Jweb_Guru 39h ago on HN
To borrow the usual quote about ingenious fools, the trouble with "easily" disproving that False is in your context is that there are such ingenious proofs of False. One can appear to be doing very substantial and sophisticated mathematics, encountering many obstacles along the way that look identical to what you'd face if you were relying on something true, only to realize much later that one of your hypotheses is uninhabited. The concern was that were the Riemann Hypothesis false, a very, very large subset of modern mathematics might fall into this category.
thevalueisai 40h ago on HN
The value is proving AI can be useful after all (and we're fucked).

Remember, not too long ago we had hordes of folks here proclaiming machines can't think, reason or be creative in any productive sense. Many still do, but it's starting to look bleaker by the day.

agentultra 40h ago on HN
This is a competent person using a tool in their work.

The tool isn’t doing the thinking or the proof on its own.

It also hasn’t been vetted/verified yet. There’s still work to do.

0c3ca83 38h ago on HN
What were the prompts?

For many of the proofs coming out of the AI companies where the prompts were posted, they ended up being variants of "solve the problem", "keep going", "I believe in you".

Many of the results could have come from a 15 year old that heard of the problem in passing.

The result here is a quarter of a million pages. Human involvement is going to be minimal.

what 37h ago on HN
If you believe these were the prompts, I’ve got a bridge to sell you.
ToucanLoucan 40h ago on HN
> Remember, not too long ago we had hordes of folks here proclaiming machines can't think

Machines can absolutely think. It's really all they do.

> reason

Murky, but LLMs prove with an avalanche of statistical modeling, they can do something kind of analogous to reasoning, with the caveat that whatever they output will necessarily have to exist in some fashion in their training data.

> or be creative

Absolutely not. By definition an LLM cannot create original work, it can only remix and recombine what it already knows. That can be interesting, and bloody useful especially for code tasks, but it is not making anything original.

retsibsi 40h ago on HN
> Absolutely not. By definition an LLM cannot create original work

What (if anything) unambiguously counts as original work, according to this definition?

ToucanLoucan 40h ago on HN
Arguing the definition of original is pointless for the exact reason you point out: that it's ambiguous. Due to the nature of how an LLM works, it cannot make anything truly new, it, again, can only recombine and remix things from it's training data. This is why AI images/writing are so easily spotted, and even AI code to a degree: they have a distinct "style" as a consequence of how they work.

I leave what is truly original to the philosophers. What I can say for certain, is that anything you get out of any of these models is a remix of what is present in it's training data, because gravity exists and energy is conserved. They are not intelligent, they are not conscious, and they cannot make New things.

retsibsi 39h ago on HN
> What I can say for certain, is that anything you get out of any of these models is a remix of what is present in it's training data, because gravity exists and energy is conserved.

But doesn't that same argument apply to humans, where 'training data' = all our sensory inputs (plus what's encoded in our DNA, depending on whether we want to analogise that to the training data or the underlying architecture)?

> Arguing the definition of original is pointless for the exact reason you point out: that it's ambiguous.

I'm not trying to be a smartarse here but you are the one who is claiming absolute certainty that something is false 'by definition', so I think the onus really is on you to have a precise definition in mind.

> They are not intelligent, they are not conscious, and they cannot make New things.

I'm not a consciousness denier or anything, in fact I think it is the most important thing in the world, but I don't understand the connections you're making here. 'Intelligent', like 'original', can be defined and needs to be defined for this statement to be meaningful. 'Conscious' is probably inherently slippery, but if we mean the same thing by it (having subjective experience/qualia) then I have no idea how you could be certain that any other entity lacks it. And it seems like you may be implying that consciousness is a necessary condition for originality, but I don't think you've explained how.

teiferer 34h ago on HN
> But doesn't that same argument apply to humans, where 'training data' = all our sensory inputs (plus what's encoded in our DNA, depending on whether we want to analogise that to the training data or the underlying architecture)?

That is true if your model of the human mind is akin to a machine model you also use to describe today's computers. But that is a huge assumption, making the argument somewhat circular. We don't really understand the human mind and therefore don't know if there could be other models for it which are not equivalent.

hardbass 24h ago on HN
If you think the human body is subject to special laws no other object in the universe is, feel free to present your evidence. No model of the human brain can exist that violates the conservation of energy and thermodynamic laws regarding entropy and information. If you have one, feel free to present your groundbreaking results.

If there has to be something breaking the current known physical laws, the best chance to find it would be in extreme conditions like black holes or stellar cores not in close to STP conditions like the human body.

WarmWash 39h ago on HN
Everything is either deterministic, random, or a mix of that.

New is just a mix of deterministic and random. LLMs are not deterministic.

You don't even need an LLM to create something new and improved. Evolutionary algorithms have been doing it since the 80's.

teiferer 34h ago on HN
> Everything is either deterministic, random, or a mix of that.

What makes you believe that?

> New is just a mix of deterministic and random.

... or that?

WarmWash 33h ago on HN
Because there is nothing else.

Every action ever in universe is the effect of either something deterministic or random. There are no causes or effects that can't be put into one of those bins.

The only "out" here would be deterministic actions that originate outside the universe, but we have no evidence of that. It also doesn't escape the deterministic/random duality.

teiferer 24h ago on HN
"There is nothing else" is a very common belief to be upended when our understanding of the world improves.

I'm not saying you are wrong but I'm trying to be careful with such absolute statements.

hardbass 24h ago on HN
Nobody is claiming new forms of energy or phenomenon can't be found. As of yet though most such were found in more and more extreme conditions, starting from radioactive substances to particle colliders and astronomical phenomenon. There is a much lower chance to believe something fundamentally or radically new phenomenon in physics to emerge from human bodies.
jonahx 39h ago on HN
> Due to the nature of how an LLM works, it cannot make anything truly new, it, again, can only recombine and remix things from it's training data

Fwiw, David Hume made this exact argument about human imagination:

> But though our thought seems to possess this unbounded liberty, we shall find, upon a nearer examination, that it is really confined within very narrow limits, and that all this creative power of the mind amounts to no more than the faculty of compounding, transposing, augmenting, or diminishing the materials afforded us by the senses and experience. [David Hume — An Enquiry Concerning Human Understanding (1748), Section II, Paragraph 5]

hardbass 39h ago on HN
Non answers are best ignored. You have failed to define what creativity is and therefore failed to prove whether llms are "creative" or not.

What exactly are you trying to say about gravity and thermodynamics? You may be confused but these are universal physical laws that apply to everything, humans, computers, anything else.

At its most basic level if I had to start with defining creativity, it should be some kind of intelligent filtering process. The brain can be in a huge number of state configurations, the interesting and novel among them are "creative" ideas. We can build upon it further. I am happy to say I am no philosopher and I will defer to those with expertise on it. But if you don't have any definition for it, its quite strange to make up some beliefs about it without proof.

drdeca 39h ago on HN
You’re the one who said “ By definition an LLM cannot create original work”, so I think responding asking for the definitions you are talking about is quite justified?

Often when someone says that something is true by definition, their claim is incorrect.

CamperBob2 38h ago on HN
Arguing the definition of original is pointless for the exact reason you point out: that it's ambiguous.

Well, it certainly wasn't "ambiguous" two posts ago when you said "Absolutely not. By definition an LLM cannot create original work." Make up your mind.

You've been beating this drum since roughly GPT 3.5, as I recall. Your goalposts aren't even on the field anymore, they're gathering dust in a far corner of the parking garage down the street from the stadium. Time to update your priors.

brookst 38h ago on HN
Does your same argument hold for humans’ inability to do more than remix? Gravity exists and energy is conserved for us, too.
haickernews 39h ago on HN
If you can be persuaded, looking into how modern LLMs are trained with reinforcement learning may change your views.
oedemis 38h ago on HN
https://arxiv.org/abs/2510.07364 event rl isn't necessary for LLMs to develop reasoning capabilities in some cases
ThrowawayR2 39h ago on HN
These AI bros don't get it. These LLM systems solving math problems have the benefit of being constrained by formal rules of math providing constant feedback when they break the rules that prevents them from wandering into la-la land. As Dijkstra said in his prescient "On the Foolishness of 'Natural Language Programming'":

> "The virtue of formal texts is that their manipulations, in order to be legitimate, need to satisfy only a few simple rules; they are, when you come to think of it, an amazingly effective tool for ruling out all sorts of nonsense that, when we use our native tongues, are almost impossible to avoid."

To whatever extent that LLMs can "think, reason, or be creative" in generating these proofs (or writing code), they're aided by very high guardrails so it doesn't say very much about what they can do in broader, fuzzier domains.

brookst 38h ago on HN
Would you say that the early results of AI beating humans at backgammon and checkers had any predictive power towards chess and go?
ThrowawayR2 38h ago on HN
The kind of AI used for backgammon/checkers and the kind of AI used for chess/go, and LLMs differ greatly so I'm not sure what the point of your question even is. A casual check says that the state of the art LLMs have much lower [EDIT]Elo ratings for checkers than the purpose built "AI" that beat humans.
Y_Y 38h ago on HN
Elo
ThrowawayR2 37h ago on HN
Thanks, I fixed it.
randomNumber7 37h ago on HN
Goedel proofed that you can formulate nonsensical statements in math (and any formal system that is complex enough to be useful)
ThrowawayR2 37h ago on HN
Gödel proved that statements can be created that are definitely true or false but for which no proof can be constructed. It is not relevant here since proofs are successfully being constructed, according to checking using Lean, by an LLM.
lern_too_spel 40h ago on HN
The post mentions some generally useful machinery derived on the way to the proof.
SubiculumCode 40h ago on HN
"... but what I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater."
applicative 40h ago on HN
That there are so many results beginning 'if RH' has exactly the opposite significance you are trying to suggest.
DennisP 40h ago on HN
There are also lots of results proven "if the exponential time hypothesis is true," and now it looks like an AI has proven it false.
ncallaway 40h ago on HN
Well, for one all those other results that assumed Riemann was true are now also known to be true. That seems valuable
baq 39h ago on HN
150GB of lean must be worth something
QuesnayJr 38h ago on HN
Doesn't that increase the value? We now know all of those results are true, where before we didn't.
kypro 37h ago on HN
If value here is a measure of utility, its not super useful on its own that something that was almost universally assumed to be true is now known to be true.

But with a problem as difficult and fundamental as the Riemann Hypothesis the thought has always been that in the process of solving it we'd be likely to develop or discover something that did have real world utility. At least this is often what happened in the past, the obvious example here being Newton developing calculus, a field of mathematics which can be applied to other problems beyond the narrow set of problems he was trying to solve.

What I find interesting about this, if the claim holds up, is whether this will assumption will remain true, or if the value of these AI proofs is going to significantly lower than those of human discoveries in the past.

Maybe a good analogy here is to think about how all of tools we humans have created has been ultimately in the pursuit of solving some problem. For the developers here, consider this – if ChatGPT had existed before React, would React ever have been developed? I mean, if an LLM can brute force the same result with just vanilla JS the need to develop frameworks and libraries to solve what was once difficult for a human to solve drops.

I'll note that I'm playing devils advocate here as I think AIs would probably build something better than React had it not existed, but it's something to think about at least. I suspect there's a higher probability than some here are assigning that this proof ultimately ends up have relatively little utility.

QuesnayJr 37h ago on HN
It wasn't assumed it was true. People thought a world in which it was true was more beautiful than one where it wasn't, but proving it true wasn't a pesky little detail.

In fact, there was a whole literature of papers proving that if the GRH was false in one particular way (the existence of Landau-Siegel zeroes) many theorems followed from it. For example, if there's a Landau-Siegal zero, the twin primes conjecture is true.

kypro 34h ago on HN
> It wasn't assumed it was true

Are you claiming that there wasn't a significant majority of mathematicians who believed it true? That's all I'm saying. I can only assume we agree.

To expand on what I said though, if it is proven that P != NP then nothing really changes as that assumption is in line with everything we understand about computational complexity theory. The proof would really just suggest there's probably nothing that surprising to be found that we don't know.

If P = NP however that is far more likely to suggest we're missing something fundamental which could radical consequences for more practical fields like comp-sci. Similar can be said of GRH. Everything we know points in the direction of GRH being true, and while it may be interesting to prove these things, its much less likely to lead to some surprising and impactful downstream discoveries.

QuesnayJr 21h ago on HN
I doubt a majority of all mathematicians ever believed it to be true. (It's hard to have an opinion outside of your area of expertise, other than "I guess this expert sounds credible.") Probably the majority of all number theorists thought it was true. But this is mathematics, not democracy.

But that doesn't change the value of proving it. Most graph theorists probably thought the 4-color theorem was true, but it was still a big deal when a computer proof showed it. Most people thought the Kepler conjecture was true, but it was still a big deal when Hales reduced it to a big minimization problem that he solved.

On the flip side, everybody thought the Hauptvermutung was true, and probably subconsciously used it. But it's not.

gauze 40h ago on HN
Even after all the MPs are beat, there's still work to do to actually understand them, communicate them, and transform them in a way that can be leveraged. I predict that's going to be much harder for AI to solve.
teekert 41h ago on HN
I think he worked together with this person? https://www.youtube.com/watch?v=Mzbtj5nkMXI

Pretty crazy interview, he didn't want throw the Anthropic employee under the bus when OpenAI offered co-authorship on the Navier Stokes paper. He suspects OpenAI somehow used his codex usage btw. Almost unprovable of course.

simonw 41h ago on HN
Yes, it was a joint effort with Tristan. Part of the scandal around it came when OpenAI offered to let Tristan help write up their result but wouldn't allow Levent to be involved because he worked for their competitor.
irthomasthomas 40h ago on HN
Choice quotes from my notes on the affair:

  To: Buckmaster
  From: A friend at OpenAI
  Subject: Openai is about to do something really stupid.
"I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, 'Why would you ruin your career?' I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, 'If you don't want me to be nice, then I don't have to be nice.'" (Sébastien Bubeck later apologized on X for the remark.) "Sebastien twice asserted that he wanted Levent removed from authorship..." "The second was that, after posting Euler, I alone write a paper presenting the Navier-Stokes result, acknowledging that an internal OpenAI model had resolved it."

"I asked whether the model had been trained on, or had access to, our sessions in Codex, into which we had been putting all our drafts for the whole of this project. I was told the model did not look up user data. I asked again, about training, and I did not get an answer." (OpenAI subsequently stated publicly that while unlikely, it could not rule out that de-identified data from their product usage improved the model.)

keeda 39h ago on HN
In broader context the “being nice” part seems like both of them talking past each other.

What happened was the two researchers, Buckmaster and Alpoge (from TFA) solved a significant subset of Navier Stokes (referenced as “Euler”) which set off rumors.

Through whatever game of Chinese Whispers ensued, OpenAI thought they had solved the full Navier Stokes problem, which they hadn’t. So OpenAI pointed their latest unreleased model at it, and it actually solved the full problem!!

Pure George Danzig vibes on that one.

So when this meeting happened, the OpenAI folks were surprised that these researchers had NOT solved the full problem, only the subset. Yet from their perspective, they were offering them credit for the full problem, and so “we’re being nice.” The Anthropic employee (Alpoge) was clearly an optics issue and so they felt they didn’t have to be that nice.

But the other researcher, Buckmaster, was suspicious that their work had been stolen, and so was understandably in a combative mood. You can see that from his perspective they were not being “nice”, they were trying to scoop them, possibly on stolen work, and then bribe them with shared credit.

But they had not solved the full problem, and academic plagiarism is a very serious allegation that can ruin careers of both the accused and the accuser. So from OpenAI’s perspective it was like, why would you accuse us of plagiarism when you yourself have not solved the full problem and when we’re willing to share undue credit for it??

This was made worse by the fact that OpenAI would not categorically deny if the others’ work had been used for training until much much later. This was terrible optics, but anyone who’s dealt with large scale data pipelines compounded by privacy issues knows that it gets very difficult to trace these things, so it takes time to make a definitive statement.

Alpoge somehow seemed less troubled by this whole episode, but I guess it’s because not only was he gonna be a multi-millionaire soon, he had Riemann stewing on his backburner.

I feel like when they finally make a movie out of all this, this scene would end up being a bit of comic relief.

ameliaquining 39h ago on HN
I think it's clear enough that something broadly like this is right, even if people don't want to admit it because the vibes of extending any charity at all to frontier labs are bad.
irthomasthomas 39h ago on HN
That version of events skips some central facts from Buckmaster’s statement.

"they came prepared and they wanted to show us the prompt... they showed me on the screen and they just literally copy pasted the Millennium Prize problem, so they said that they took the prize and they pressed enter and they got a solution, and at that point I just had to laugh because it's just so ridiculous. It came out within a few minutes that this was a lie"

"Levent had been told by Sebastien 'very little human input' had been used. This turned out not to be true. Over the course of the call, as members of their team sent Sebastien corrections and details over their internal chat, it emerged that an entire team had been working on the problem..."

"I was offered the opportunity to write as sole author of their paper and throw my collaborator under the bus... they were asking me to throw him under the bus, someone who I've been working with for a year, to throw him under the bus just because he's an Anthropic employee"

"the right paper to compare is the unforced Euler equation. Now the unforced Euler equation, all the mechanisms in this unforced Euler equation are identical to a different version of the Euler paper that we haven't released yet... basically all the main architectural ingredients in that unforced Euler were in our work for which they would have had on their servers, and it hadn't been published" "they didn't use 10,000 agents to get to the Euler problem, they used 100 agents. So they worked from absolutely nothing with only 100 agents... and then they ended up with an identical architecture to what we had"

keeda 37h ago on HN
I don’t see how those details change the overall outline of what I presented above through? Maybe you could nitpick that it was a team with a model using 100 agents, and not a solo model with 1000 agents. But the fact remains OpenAI ended up solving the bigger problem themselves just because they thought the researchers had solved it already when they hadn’t. That is low-key hilarious.

The Euler plagiarism insinuations are much more serious but given all the 100s of other results we’ve seen come out, highly unlikely.

randysalami 37h ago on HN
Not to put anyone on an idol but the only one who comes out of this not acting like a greedy, anti-human is Buckmaster and his colleague (I think). Straight up lying by OpenAI, very predatory "meritocratic" behavior. What gets me the most is "they didn't use 10,000 agents to get to the Euler problem, they used 100 agents. So they worked from absolutely nothing with only 100 agents... and then they ended up with an identical architecture to what we had". What the hell! So they lie about their technical claims too on a factor of 100x?? And companies are taking their words seriously, no wonder there is a house of cards being built in the industry.
bdelmas 38h ago on HN
Terrible optic is such a bs executive wording when they lied, coerced, bullied, and threatened him…

I like the summary but knowing what irthomasthomas posted that’s missing from it, and having a better picture of what happened; terrible optic doesn’t work.

freejazz 32h ago on HN
>In broader context the “being nice” part seems like both of them talking past each other.

Really?

keeda 29h ago on HN
Yes, because when looked at from the outside, like I tried to do with my recap, it's clear that:

1) OpenAI thought they were "being nice" by still offering (partial) first authorship despite having independently solved a bigger problem than the researchers.

2) But Buckmaster thought OpenAI were "being nice" as a means of CYA / bribing / coercion for scooping and plagiarizing them.

"Being nice" has very different connotations in the two contexts.

freejazz 28h ago on HN
Thinking you're nice doesn't mean you're actually nice. OpenAI isn't talking past anyone. It's talking out of it's ass.

Buckmaster didn't think they were being nice. He thought they were being assholes to him and trying to bribe him.

Putting the phrase "in quotes" to use it completely out of context, only to make the point that surely no one understood each other... that's some impressive work. If you have to work this hard to make sense out of a conversation, it's because you're not.

keeda 19h ago on HN
Offering first authorship on a significant paper to a researcher who had not even solved the whole problem seems a lot like being nice (not "in quotes" ;-)) to me too, but that's just me.
freejazz 14h ago on HN
You are again ignoring the context that is offered as part of a bribe. I don't understand why you keep ignoring that context.
keeda 10h ago on HN
But I just keep missing that the bribe was offered for what exactly? For accidentally solving a bigger problem than they had?

If I unintentionally upstaged somebody by achieving a greater feat I would have empathy for them, but I would not feel the need to bribe them. What would I even be covering up here?

Really the only insinuation worth covering up is if the work had made it into their training data, but the hundreds of proofs that have been released subsequently lend significant weight to OpenAI’s categorical denial of that possibility. I also think it would be a company-ending move, akin to Google mining your Google Docs for trade secrets. OpenAI cannot afford this scandal when the Enterprise race is so cutthroat and switching costs are 0.

FWIW I’ve posted before that IP is likely one of the few remaining moats for smaller players, so I am pretty paranoid about this issue. I’m working on research-y things myself using AI (because I can’t afford to hire a PhD!) that I hope to protect and capitalize on. I compulsively check the “training opt out” settings every so often. So I would really like to know if I’m missing something.

freejazz 3h ago on HN
>But I just keep missing that the bribe was offered for what exactly? For accidentally solving a bigger problem than they had?

No, for staying silent, which he chose not to do. This isn't a hard story to follow.

>If I unintentionally upstaged somebody by achieving a greater feat I would have empathy for them, but I would not feel the need to bribe them

Frankly, you don't seem to have normal social interactions at all so I'm not sure what value your feelings are here.

zelphirkalt 37h ago on HN
"While unlikely ... it could not be ruled out ..."

So basically he hit the nail on the head and they want to lie about it without lying. Either that, or they are supremely incompetent, not knowing even what they ingest into their models. Both are very much not inspiring any faith in their trustworthiness.

Betelbuddy 37h ago on HN
>> Almost unprovable of course.

Very much provable, if he would sue, and a lawyer would ask for discovery.

BoppreH 37h ago on HN
He was also the one who dropped the Jacobian conjecture counterexample on twitter with the famous

> hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

and prompted Fable with

> You should do a breakthrough

keybored 42h ago on HN
* Levent Alpöge
gus_massa 40h ago on HN
(I think you posted this comment because the parent comment has a typo that was fixed later. Let's hope it goes back to black soon.)
jascha_eng 41h ago on HN
I mean a single person can always make mistakes regardless of how smart they are. But this does make it sound a lot more exciting than if a random vibe-prover came up with it.
jrootabega 41h ago on HN
No theycan;'t
niksmather 41h ago on HN
Brother it might be worth looking up Andrew Wiles
igleria 41h ago on HN
Congratulations on making my palms sweaty. Fields medal for Levent/Claude...
talon8635 41h ago on HN
You mean, if the AI says it does, than it does. And it’s probably right, because it’s smarter than all of us and we basically just have to trust it now, which is exciting and frighting in many ways.
fearmerchant 41h ago on HN
Thanks, the title made it sound like the custodian walked by and typed "solve this" and found an answer.
stavros 41h ago on HN
Man, Good Will Hunting 2 is much more boring than the original.
rfgplk 41h ago on HN
That's pretty much what he did regardless. I can assert with absolute confidence that he read effectively 0% of the "proof".
hersko 41h ago on HN
How could you possibly know that?
svnt 41h ago on HN
Because the writeup describes exactly that process:

> Jarred Sumner, an Anthropic staff member (and non-mathematician), prompted Claude to “take a real stab” at the hypothesis itself, leaving the mathematical choices from there up to the model. Initially, Claude generated and tried 650 ideas, none of which worked. Jarred prompted Claude to try again, and it spent a day and a half coordinating about 60 Claude subagents, which this time went much deeper: between them, they ran 2,400 shell commands and wrote hundreds of Python scripts.1 The subagents ran thousands of numerical checks against known zeta zeros and refereed one another’s work. Throughout this process, Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”).2 This seems to have helped Claude overcome some initial skepticism that it could make meaningful progress.

> Having found this new result while attempting the task, Claude tested its work by having various subagents review the proofs, search for counterexamples, download 54 papers from the arXiv to check that its finding hadn’t already been made, and independently re-prove its finding from scratch. Claude volunteered to write its findings up as a paper, and recommended that a human number theorist validate its findings.

> Levent Alpöge and Ralph Furman, two of Anthropic’s own mathematicians, examined Claude’s work to understand the new results and how they related to the prior work mentioned above. In parallel, Claude worked with another member of staff, Eric Easley, to produce a Lean formalization of the result, which passes the standard validation tool comparator.

jasonlotito 40h ago on HN
Your comment literally disproves your own comment.

Simply put:

> Because the writeup describes exactly that process:

This is a complete lie.

svnt 40h ago on HN
He told to find a solution. It failed. He told it to try again, and to keep going when it stopped.

He had no mathematical skill or ability to check the solution.

so which part is the complete lie?

jasonlotito 40h ago on HN
The third paragraph. You left that entire part out: the actual proof, the actual confirmation. The thing that's literally linked to at the top of this thread.

There's an entire process here, and you just decided to leave the most important part out. So, your comment literally disproves your claim. Which makes you a liar. Go ahead. Get to this thread by removing that third paragraph. That is the critical part of this entire thing.

svnt 39h ago on HN
You are conflating:

a) “looked at claude’s work”

b) read and understood over 200,000 pages of ai-generated proof

everything they post to their website is marketing

You also seem to be conflating “Levent guided the work” with “curious employee brute-forced some potential solution through an agent tree”. He (in this case both Levent and said employee) did not guide the work, it was in effect a random employee saying “solve this” over and over.

mkarrmann 40h ago on HN
It's a 250,000 page proof. By his timeline he had the first breakthrough about a month ago, and he admits to not understanding everything yet.

Even if he were extremely prolific at reading through the math, he would have carefully read <1% of the proof in the last month.

_fizz_buzz_ 41h ago on HN
Huh, what a conincidence that the AI presented the proof to Levent Alpöge (a top class mathematician) and not to some Rando McRandomface then? What are the odds?
svnt 41h ago on HN
I posted below in detail, but the answer is because Levent works there, not because Levent “guided” anything.
_fizz_buzz_ 37h ago on HN
He doesn’t really go into much detail here, but according to him he came up with the strategy: „ I had us on the line of combining two quite different ideas present in a toy setting, one of which was already new and shocking, making very simple one of the shadows of Riemann over which we mathematicians felt most triumph. I guided us to the right situation analogous to that toy setting, and then Claude began to leave me in awe.“

That does not sound like he typed: „solve rieman thingie, make no mistakes, plz!“

svnt 36h ago on HN
Here is the description of the initial process:

> Jarred Sumner, an Anthropic staff member (and non-mathematician), prompted Claude to “take a real stab” at the hypothesis itself, leaving the mathematical choices from there up to the model. Initially, Claude generated and tried 650 ideas, none of which worked. Jarred prompted Claude to try again, and it spent a day and a half coordinating about 60 Claude subagents, which this time went much deeper: between them, they ran 2,400 shell commands and wrote hundreds of Python scripts.1 The subagents ran thousands of numerical checks against known zeta zeros and refereed one another’s work. Throughout this process, Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”).2 This seems to have helped Claude overcome some initial skepticism that it could make meaningful progress.

Only after this person thought he/claude found a solution were the mathematicians brought in.

lg5689 32h ago on HN
This description is about the 67% bound from a few months ago, not the full GRH being discussed here.
svnt 9h ago on HN
And you think they are two isolated processes?
rfgplk 40h ago on HN
A Rando McRandomface doesn't have access to millions to tens of millions of dollars to spend on tokens.
rcxdude 39h ago on HN
There's a lot of Random McRandomfaces in Anthropic with the same access.
steve_adams_86 40h ago on HN
This isn’t anything like what using LLMs for complex problems looks like.

I sometimes think AI doomers are terrified because they have no idea how this all works.

sebzim4500 39h ago on HN
It isn't how this work was done, but it is basically how UGC and 400 other high profile theorems were proved by OpenAI so I'd say the doomers have a point.
xyzsparetimexyz 41h ago on HN
What do I have to do to get an asteroid named after me?
amenghra 41h ago on HN
Change your name to 354 Eleonora ¯\_(ツ)_/¯
block_dagger 37h ago on HN
Wouldn't that be naming oneself after an asteroid? Not quite what the OP is after.
heisgone 41h ago on HN
Let's redefine AGI as solving every known problems to man so we can stop moving the goalposts.
yesbut 41h ago on HN
it hasn't solved any of mankind's problems yet.
DFHippie 41h ago on HN
It will need cooperation from mankind. So do we.
aeonik 41h ago on HN
It's solving tons of my problems every day...
win311fwg 41h ago on HN
Hence why the previous comment expressed "mankind", not "HN accounts".
stale2002 41h ago on HN
I think the point here is that quite a lot of problems are being solved and this is entirely self evident based on the stuff people are using it for right now.
derektank 40h ago on HN
What are “mankind”’s problems if not the problems of individual humans? It seems like we’re treading towards a definition of AGI that would imply most human beings aren’t generally intelligent, which seems wrong headed.
win311fwg 39h ago on HN
Mankind's problems are that of humans but not that of software.
toomanyrichies 34h ago on HN
It would really be something if AI could fix humanity's tendency to confuse "my lived experience" with "mankind as a whole".
bhaak 40h ago on HN
Have you considered that you might not be representative of humankind?
aeonik 38h ago on HN
I have actually considered the fact that I may secretly be an alien embedded into humanity, or perhaps a dog in disguise.

But evidence really points to the contrary, so... no, I'm pretty confident I'm human with regular human stuff that AI solves for me. House and family stuff on top of the science and tech stuff that applies to a smaller portion of human kind.

apical_dendrite 40h ago on HN
It's certainly creating a lot of new ones!
lionkor 41h ago on HN
You can have the AGI label, you can put it next to the "actually, we've solved world hunger, people just aren't working hard enough" discovery, somewhere on the shelf with "every billion I make trickles down to the barista when I get a coffee at starbucks".

The AGI label doesn't matter. You can call it AGI when you feel like it's generally intelligent--I've yet to see any model that even remotely seems intelligent, but it all depends on where your bar is.

If this is superintelligence to you, that doesn't make it superintelligence to me. I can not have a single chat with an LLM without running into it clearly just predicting whatever sounds good instead of actually reasoning. I can't understand what sort of work people are doing where this seems super intelligent, or intelligent at all.

sebzim4500 40h ago on HN
>I can't understand what sort of work people are doing where this seems super intelligent, or intelligent at all.

Proving the Riemann Hypothesis apparently

lionkor 25h ago on HN
But not in some sort of super clean elegant way, it's "just" brute force, no? The same way it brute forces fixing tests until the software works, instead of designing coherent software in the first place.
kolinko 40h ago on HN
What is your definition of intelligence?
thevalueisai 40h ago on HN
Proving GRH never was the hard part of the job anyway..
keybrd-intrrpt 40h ago on HN
How about when it picks me, and nobody else, the winning Powerball numbers.
GuB-42 39h ago on HN
And yet, we still can't make a robot that can clean a room. I mean really clean a room, picking up trash and put it into the bin, clothes on the floor go into the laundry basket, remove stuff from the shelves, dust it and put it back, etc... the kind of job any maid can do. Not like a Roomba that can be defeated by a loose sock.

AGI means "general", a machine that can solve problems geniuses have struggled with while failing at problems that can be solved by minimum wage workers is hardly general, instead, it is a good specialized tool.

012forthee 39h ago on HN
If Moravec's paradox is anything to go by, the "embodied" tasks will end up being the ones most difficult for AIs to do.
croes 41h ago on HN
> If he says he has Riemann, well then, he has Riemann.

That’s not how this works, especially if he works for Anthropic

artninja1988 41h ago on HN
Is it really wise to only rely on credentiallism to verify things? Plenty of mathmaticians top of their fields believed they have proven something when the proof contained some flaw. Just look at Mochizuki
317070 41h ago on HN
credentialism, and a lean codebase that compiles and certifies Riemann is correct.
well_ackshually 41h ago on HN
No. A lean codebase that certifies that whatever definition of the Riemann hypothesis their LLM wrote is correct.

I can write Lean that certifies that 2 + 2 = 5 is correct, by redefining every axiom about natural numbers on the way. Or abusing bugs in Lean. Stop believing Lean is the response to everything.

latent-person 41h ago on HN
Just like Navier stokes etc they use independently written Lean statements written by humans; formal conjectures [1]. Indeed, they have one for the Generalized Riemann Hypothesis too [2].

> I can write Lean that certifies that 2 + 2 = 5 is correct, by redefining every axiom about natural numbers on the way.

Which you would catch using comparator?

> Or abusing bugs in Lean.

And just like all the other ones, people check with multiple independent kernels, so you need to abuse bug(s) in all of them.

> Stop believing Lean is the response to everything.

Stop believing Lean is so flawed.

[1]: https://google-deepmind.github.io/formal-conjectures/

[2]: https://google-deepmind.github.io/formal-conjectures/theorem...

astrange 41h ago on HN
It imports a standard definition of RiemannHypothesis. It could be wrong in other ways, but not that one.
steinwinde 41h ago on HN
>whatever definition of the Riemann hypothesis their LLM wrote

Given that the proof is provided by Levent Alpöge, who is known to be more than familiar with Lean, is it not unlikely he made up his own definition of the problem?

https://leanprover-community.github.io/mathlib4_docs/Mathlib...

stale2002 41h ago on HN
> whatever definition of the Riemann hypothesis

Do you... think that there might be existing definitions for one of the most famous math problems of all time and that maybe they used that?

Or do you think that they just vibe coded up a different definition, for kicks?

This kinda feels.... easily provable either way.

quchen 41h ago on HN
The axioms would be a given to the theorem prover of course, as is the norm for theorem proving. This then reduces a complex proof to checking whether the theorem is stated correctly and the proof kernel is correct. This may not be (provably) easier, but it is so far removed from yet equivalent to the original problem that you get a machine-verifiable independent audit as a result.

And don't reduce "is not the answer to everything" to "is the answer to nothig", there is tremendous value in automated verification, especially in the AI age where the moat to it going mainstream is shrinking by the day.

math_dandy 40h ago on HN
The statement of the Riemann hypothesis is included in the Mathlib library which is a human written and highly regarded foundation for formalization. See https://github.com/leanprover-community/mathlib4/blob/master... for the two-line statement.

For a discussion of how to check that the untrusted proof isn't playing fast-and-loose with the axioms of mathematics or exploiting kernel bugs, see Kevin Buzzard's recent post "...and the proof is checked in Lean" at https://xenaproject.wordpress.com/2026/11/12/and-the-proof-i....

questionable_ju 41h ago on HN
he's just pointing out that this isn't some rando claiming to have video of aliens.
lanstin 41h ago on HN
Math is hard; even to understand the questions and what the entities are which the questions address takes a lot of deep study, even for smart people. So credentials are not sufficient but they are for sure necessary. I have tried my whole life as a software engineer to study math as an amateur and then started a masters program and learned more details thoroughly in three months than in decades of self teaching.
hephaes7us 39h ago on HN
The study is what's necessary. The guidance of a program is a helpful accelerant. A credential is merely suggestive.
lanstin 39h ago on HN
Not in a Bayesian sense, not "merely" suggestive, but quite suggestive. Not sure how many full professors you have met, but they aren't like that under-performing coworker that people have to cover for.

More like IIT grads in being, Bayesianly confident, quite exceptional.

Davidzheng 39h ago on HN
Why is it necessary--just check the proof
teiferer 34h ago on HN
> I have tried my whole life as a software engineer to study math as an amateur and then started a masters program and learned more details thoroughly in three months than in decades of self teaching

Oh man, you can't imagine how much I'm cheering you on and rooting for you. I wished more peope could embrace that insight/approach.

dgellow 41h ago on HN
No it’s not. HN main approach to analyzing anything is to rely on chaining heuristics. Which is generally sort of fine for software, but doesn’t make sense in the real world. It’s at best one signal to weight in your analysis, nothing else
wat10000 41h ago on HN
It's the only realistic way to operate for nearly everything. Try determining whether a sandwich is safe to eat without applying a shitload of heuristics.
swatcoder 40h ago on HN
The difference is that you generally need to eat something and so drawing some conclusion about th sandwich is essentially forced upon you.

Getting buzzed over headlines about math or medicine or whatever, on the other hand, is just an irrelevant hobby of bored gentry and the opportinity to just not draw a conclusion at all remains open.

wat10000 40h ago on HN
Do you just not contemplate anything that isn't directly related to your physical needs? What a sad existence. I wonder what you're doing here, in that case.
TeMPOraL 39h ago on HN
Tell that to the general population wrt. politics. The "opportunity to just not draw a conclusion" has a much greater payoff there.
mtlmtlmtlmtl 40h ago on HN
Easy. Eat the sandwich, see what happens. No heuristics necessary.
killerstorm 40h ago on HN
Nobody claims it's verified. It just seems very plausible.

It really irks me that people think beliefs have to be binary rather than come in different degrees of certainty.

randomNumber7 37h ago on HN
This just depends on intelligence.
teiferer 34h ago on HN
Given their typical sky-high confidence, does that prove that AI agents are not actually intelligent?

scnr

killerstorm 22h ago on HN
Not at all. A person doesn't need high intelligence to understand what "maybe" / "it might be true" means.

It's just pro-science try-hards who believe that things can be verified by science to become facts, absolute truths, etc

randomNumber7 21h ago on HN
I would honestly like to know what you think is the difference between those groups of people. I thought about that a lot and couldn't find a better explanation.

Why do some people understand what "maybe" / "it might be true" means, while others believe in absolute truths?

EGreg 40h ago on HN
I heard Fermat himself had a marvellous proof, but it was too small to fit into the margins. A pity!
abnry 40h ago on HN
I have a marvelous lean proof but my hard drive is too small to contain it.
EGreg 3h ago on HN
Doesn’t seem very lean, actually…
tptacek 39h ago on HN
Hacker News is a water cooler conversation, not a peer review system. It is in fact quite useful in this context to short circuit a whole raft of standard dismissals!

We went through this with Alpöge over the Jacobian Conjecture and the resulting thread was pretty embarrassing.

moralestapia 39h ago on HN
On this water cooler conversation, I sure appreciate the comments of my fellow @artninja1988.

If anything goes, anything goes.

supern0va 38h ago on HN
> If anything goes, anything goes.

This sort of black and white thinking isn't helpful. Their point in giving context about the author is to give you more data for your internal heuristic to figure out whether this result is plausible and worth discussing.

As opposed to, for example, someone with AI psychosis on Twitter posting that they've solved it.

Even if Terence Tao came down from the clouds with a beautiful human-written proof, no one would be certain that it is solved--people make mistakes. But again, it gives confidence and something fun to discuss until the math community digests it fully and comes to a consensus.

Anyway, TL;DR is that no one is telling you to go put your life savings on a prediction market bet that it's solved. Relax a little.

moffkalast 36h ago on HN
Can't even get complementary cold water around here these days smh.
hyperpape 38h ago on HN
No. The mathematicians will verify this, and come to a conclusion.

What the credentialism does is establish that this should be taken seriously and has a real possibility of being right. The Lean also helps (no Lean for Mochizuki’s work).

RunSet 38h ago on HN
I understand the proofs generated by the slop engines are beyond mortal kenning, such that they must be backstopped by appeals to authority.
some_furry 41h ago on HN
> If he says he has Riemann, well then, he has Riemann.

Yeah, no. This kind of framing is inappropriate. Let the mathematicians check the proof.

adastra22 40h ago on HN
That’s what he is saying though?
some_furry 40h ago on HN
Is it? It sounds to me like glazing one person in particular rather than the rough consensus of the entire mathematics community. I'm saying the latter is necessary.
rfgplk 41h ago on HN
I don't understand this comment. For instance,

> a supremely qualified mathematician

How? No matter how "qualified" you are, the AI is still doing ~100% of the work. And since we're talking about superintelligence (Levent's intelligence hasn't even been stated anywhere), it's questionable whether a human can even "verify" or "check" the AI's work.

_fizz_buzz_ 41h ago on HN
- It's a lot more straightforward (easier) to verify a proof than to come up with a proof

- It is also unclear if this proof is something a human couldn't have come up with

fooker 40h ago on HN
> It's a lot more straightforward (easier) to verify a proof than to come up with a proof

I don’t disagree, but this is not established science.

We actually don’t know if it’s easier to verify a proof than to come with a proof.

It’s one of the fun things that lead to P!=NP if it’s correct.

Again, probably correct but if you could prove it that would be groundbreaking.

HotHotLava 41h ago on HN
If you read through the pdf that this post links to, you'll find that while claude contributed many ideas and was ultimately responsible for the technical "implementation" of the proof, there was also a lot of non-trivial guidance and steering involved from Alpöge himself. He didn't just sit down and prompt "claude, solve the riemann hypothesis for me".

If it was that easy, OpenAI would have already announced the result weeks ago.

simonw 40h ago on HN
> the AI is still doing ~100% of the work

That's clearly not true here.

steve_adams_86 40h ago on HN
> the AI is still doing ~100% of the work

So if we stick a 10 year old in front of Claude and instruct them to type in “create a proof of the Riemann hypothesis”, as long as they keep pressing return, eventually a proof will pop out?

HappyPanacea 39h ago on HN
Yes because their lifespan is long enough AI will likely continue to improve
auntienomen 41h ago on HN
Alpoge is a smart and principled guy, and he is probably the world leader at getting LLMs to do math. But he's not an expert in these subjects. If we're going to argue from authority, let's wait for an actual authority to weigh in
killerstorm 40h ago on HN
Did you read the document? He is very much an expert. JFYI you don't need a medal to be an expert.
slashtom 40h ago on HN
The paper is not a proof and has not been peer reviewed / published in a journal. So yes, let's wait and see.
killerstorm 40h ago on HN
There's no paper. It's just a letter about the discovery.
auntienomen 39h ago on HN
The lean proof is public. Although, if it's anywhere near as long as the "natural language" proof, most people probably don't have enough compute to check that it compiles.

footnote: I object to calling the langauge that Claude produces "natural". It appears to be uniformly distributed over the space of grammatical constructions.

auntienomen 40h ago on HN
No, you need a long track record of working on the Riemann hypothesis. You need to be the person who's called to verify a possible proof of the Riemann hypothesis. I have a lot of respect for Alpoge, but he is not the person a journal would call to verify this.
darkwater 40h ago on HN
Asking out of real curiosity, not trying to provoke at all: and what happens when somebody verifies it and say the result is legit? Would Alpoge become an expert on the topic or wouldn't, due to the AI assistance?

Also, if it were the most expert person on the Riemann hypothesis the one working with AI to prove it, how would the verification work? The 2nd most expert person? Or would they be trusted directly?

auntienomen 39h ago on HN
It's a fun question. I don't think AI assistance transfers expertise from the agents to the user. Alpoge says much the same thing himself, noting that Claude produced 250000 pages of slop and that he doesn't trust his own summaries in places.

For the other question: We'd want multiple independent expert verifications.

akie 38h ago on HN
Man, are you being deliberately obtuse? 250000 pages of proof in a formalized language is not "slop"
auntienomen 36h ago on HN
250k pages is the "natural language proof" that Alpoge chose not to publish because it was slop.

The lean proof is 17M lines. That did get published, although good luck compiling it if you don't have access to 1/2 TB of RAM.

francasso 40h ago on HN
I don't get it, if the proof is formalized in Lean, the only expertise you need is running lean, asking if there were any additional axioms added in the proof that shouldn't be there, and if the statement being proved is the GRH (verifying all the definitions involved match expectations).
LtWorf 40h ago on HN
You need to check that there are no bugs in the lean formulation of course.
francasso 39h ago on HN
If that means checking the formalized theorem states the GRH correctly yes, but you hardly need to be an expert in the RH to do that.
vetronauta 37h ago on HN
The usefulness of a Lean proof is that it is possible to just look at the theorem statement and at the axiom that are used. Riemann hypothesis is quite straightforward to define and to be checked even by a undergrad. Obviously a bug in Lean kernel is still possible.
pell 39h ago on HN
That’s a strange form of credentialism. I understand it if you wanna reject amateurs who make graphs and don’t understand what they’re saying but this is an actual mathematician and clearly he has thought about this particular problem for a long time.
killerstorm 39h ago on HN
From the document: "It was my secret in my student days that my plans were for this problem in decades, which meant then that I tried to become familiar with all tools that I could possibly recognize in the problem when it was time".

It's true that he haven't been publicly recognized as a top expert. But it's not like he's just a software dev who got lucky with prompting - he's a mathematician who specifically studied this problem.

I don't think we should delegate gatekeeping to journals - a lot of crap is published, by "experts". (Also a lot of good results don't get published - there are well-known examples.)

xyzzyz 40h ago on HN
Of course he is an expert in these subjects. He has a mathematics PhD from Princeton, with a thesis on algebraic geometry. If you cannot call someone like that an expert, you cannot really call anyone that.
serioussecurity 40h ago on HN
Math doesn't work like that sadly.
conradkay 40h ago on HN
The model(s) that solved it is an expert, I guess. Maybe they can send it to OpenAI and have their internal model check.
auntienomen 40h ago on HN
It probably looks like that to someone outside mathematics.

But a math PhD from Princeton (even in arithmetic statistics, which is what he actually studied) doesn't make you an expert in the Riemann hypothesis.

bad_haircut72 40h ago on HN
Its all relative. If you're on hackernews then conpared to you, dear reader, hes an expert.
auntienomen 40h ago on HN
The beauty of the internet is that no one knows you're a dog.
danlitt 39h ago on HN
What an odd thing to say. By this logic dumb people should defer to worse experts because "compared to you, they're very smart". Dumb people need the best experts of all!
dang 38h ago on HN
"Please don't sneer, including at the rest of the community." It's reliably a marker of bad comments and worse threads.

https://news.ycombinator.com/newsguidelines.html

p.s. FWIW, I know of at least one Nobel Prize winner on HN.

FeepingCreature 40h ago on HN
His writeup says he's been angling to work on Riemann since he was a student. It does seem plausible he's an expert.
auntienomen 40h ago on HN
I'm going to wait for Sarnak to weigh in.
ahurmazda 40h ago on HN
Alpöge is a grand-disciple of Sarnak.

Alpöge -> Tsimerman -> Sarnak

_flag 40h ago on HN
Sarnak weighing in is completely irrelevant. There is a lean formalisation, and it either compiles or it doesn't. (And the statement of RH is well established so there is no issue there.)
nocturn9x 38h ago on HN
Lean can and does have bugs. It compiling doesn't mean anything, because that doesn't tell you whether the proof actually encodes what the mathematician intended. Plus, again, there can be prover bugs.
_flag 36h ago on HN
In principle, sure. But the likelihood of a plausible proof strategy produced by an AI agent with human assistance somehow accidentally exploiting a Lean bug is incredibly low. In any case, the fact that the Lean compiles is certainly going to be, in this setting, a more reliable assessment of the proof's correctness than the opinion of any expert, unless that opinion also comes with an identification of the bug being exploited.
Tuna-Fish 39h ago on HN
From the writeup for the lazy:

> It was my secret in my student days that my plans were for this problem in decades, which meant then that I tried to become familiar with all tools that I could possibly recognize in the problem when it was time

causal 40h ago on HN
Feels like we're reaching for arbitrary levels of expert to undermine the claim. I'm sure the world's foremost Riemann Hypothesizer will come forward at some point, but we don't need to wait for that to take the Princeton math PhD seriously.
Davidzheng 40h ago on HN
There are actually no experts on the Riemann hypothesis bc before now no one knows how to do anything lol
xyzzyz 40h ago on HN
I have a graduate degree in mathematics, with a thesis in algebraic geometry myself. I know what I know and what are my limits. I am not an expert in Riemann hypothesis now, but if I spent a couple of months of focused study, I would be. That’s where Levent surely is right now, at the very least.
auntienomen 39h ago on HN
I think we mean different things by expert. You appear to mean "superficially familiar with some known approaches to the problem". That's all you're going to get in months.

And, yes, I've also published papers in algebraic geometry.

xyzzyz 39h ago on HN
Yes, if by “expert” you mean one of the top 10 people in the world, then sure, but that not how regular people typically use the word.
jeremyjh 39h ago on HN
I disagree. I don’t know that the answer is 10, but its definitely a very small fraction of math PhDs and they all know each other.
Vachyas 39h ago on HN
If only 10 people understand the problem well enough to contribute to it, then yeah, that would be a case of a problem with only a handful of experts.

I don't think regular people would have a problem with that, since the idea that some problems could be _that_ deep isn't particularly new.

datsci_est_2015 38h ago on HN
Silly definition. I'm a "top 10 people" in the application Data Science to [niche that my company works in] because there are so few people working in this area, but I've only scratched the surface in my 3 years of working specifically in this domain.

Some things just really aren't well-understood or well-studied.

And other things are extremely well-studied and well-understood where 10 is significantly too low a limit.

godelski 39h ago on HN
You two are just talking past one another and using different, and valid, definitions of experts. Yes, Alpoge is an expert in mathematics. Though how much of an expert he is in number theory? I don't know, I'm not an expert in number theory myself, I don't know Alpoge, and I don't know where he's been focused lately.

But either way both of your arguments comes down to speculation unless you two are also experts in number theory and know Alpoge or at least what he's been working on (to a sufficient level).

fn-mote 37h ago on HN
> if I spent a couple of months of focused study, I would be

It seems more likely that in a few months you would discover the depth of what you do not know about RH.

If you are a very accomplished geometer, I apologize, but from anyone else this would be very egotistical.

justMaku 25h ago on HN
> If you are a very accomplished geometer, I apologize

He is not.

golemotron 40h ago on HN
Credentials are not authority. Especially in math.
HarHarVeryFunny 40h ago on HN
Apparently he has a Lean proof, so all that needs to be verified is that the Lean formulation of the problem matches the hypothesis, which I assume he's qualified to do (and has done).
hmokiguess 41h ago on HN
If only we stopped with the marketing and named the person, so sad to see someone like him reduced to "Anthropic Employee"
racl101 41h ago on HN
Wonder what his intellectual peers think.
gosub100 41h ago on HN
> If he says he has Riemann

Appeal to authority fallacy. The proof should stand or fall on its own.

breakwaterlabs 41h ago on HN
Appeals to authority are not inherently fallacious if the individual is, indeed, an authority on the topic at issue.

To the contrary: we lean on authority opinion in most spheres of life because one individual's ability to understand is limited and so we abstract away "true" in auxiliary fields to mean, roughly, "held to be true by trusted authorities". We even rely on it in law.

rvz 40h ago on HN
Just like when Andrew Wiles discovered the proof to fermat last theorem and Grigori Perelman with the poincaré conjecture, multiple authorities in the field got together and verified that the proofs were correct (even after Wiles initially had an error.)

An appeal to multiple trusted authorities is better than a singular authority. So at best I'll keep as unverified until it is checked, unless you have read the 250,000 proof and checked that there is no issues or soundness bug with the lean proof.

thevalueisai 40h ago on HN
They are inherently fallacious as in logically unsound. There is no connection between his credentials and him being right about this particular thing.

It's very important to keep those things separate.

breakwaterlabs 39h ago on HN

    They are inherently fallacious as in logically unsound.
At the risk of triggering an infinite regression (or logic loop), the redoubtable Stanford Encyclopedia of Philosophy's page on fallacies provides the following (https://plato.stanford.edu/entries/fallacies/):

    The ad verecundiam fallacy concerns appeals to authority or expertise. Fundamentally, the fallacy involves accepting as evidence for a proposition the pronouncement of someone who is taken to be an authority ***but is either not really an authority or a relevant authority***.
Once more, I would point you to the fact that even the legal profession-- which is fundamentally one of the most epistemologically pure fields-- spefically allows the opinion of those deemed "authorities" to be submitted as evidentiary.

See: https://www.law.cornell.edu/rules/fre/rule_702

Without an epistemology that allows integrating truth-by-authority, you by necessity need to prove every assertion from first principals. Have you done so?

interestingsir 37h ago on HN
Very interesting sir. Thank you for this exposition.

But I do not see how this changes the notion that being an authority does not turn your arguments sound or even valid. There is nothing to connect. I can be an expert and completely wrong and there is nothing epistemology can do for you there.

The legal profession is taking a shortcut like the CS profession is taking a shortcut with the halting problem. Sure, we can end the discussion that way but we cannot use it as an argument otherwise.. halting problem solved, just press control-C?

Zizizizz 36h ago on HN
He's saying that it isn't right automatically because he said so. It adds significant credibility that the proof is correct compared to me saying I did it but the proof itself is what matters (at least to most people)
gosub100 32h ago on HN
You can't even contribute to math at this level without an advanced degree (cranks and delusional people notwithstanding). So maybe I should clarify I mean "authority" as in someone with high repute among his colleagues.

If we were all factory technicians, and Max has the best record among us at fixing the machine, that doesn't mean his word is final. "Max said so" is an appeal to authority. We are all capable of reading the documentation and performing our own repairs, and sometimes Max can be wrong.

ndsipa_pomu 41h ago on HN
Yes, that's right, but proof checking does take time and effort. For an initial announcement, the fact that they are a well regarded mathematician means that it's worth taking seriously and not that it's definitely correct. If it's someone unknown, then you'd definitely be waiting for the proof to be checked before talking about it.
rvz 40h ago on HN
That is why it is still a "claim" of a proof, until other mathematicians verify it.

Not saying it is incorrect but, all claims of mathematical proofs have to be checked over formally regardless of who discovered it.

With this particular mathematician it just means it should be taken seriously but you cannot blindly believe it is correct without multiple authorities testing that there are no holes in the proof, just like they did with FLT and the poincare conjecture.

5upplied_demand 40h ago on HN
> The proof should stand or fall on its own.

At some point in the chain an "authority" has to validate whether the proof stands or falls.

soperj 40h ago on HN
sure, still wouldn't trust the guy who wrote the proof to verify it though, which is what this person is asking.
Spacecosmonaut 40h ago on HN
Of course it has to. Its not about that, its about whether or not to get excited. I think we can start to get excited.
emtel 40h ago on HN
obviously the claim is about how much credence to assign the claim until the proof has been verified
jsLavaGoat 39h ago on HN
>fallacy

It doesn't make it right, but more importantly to people who always quote every known fallacy: it doesn't make it wrong either. We shall see, empirically how it goes.

But it is highly relevant whether this is some crank on an Internet forum posting proofs about infinity-hat or someone with some credentials, even if that framing was a bit hyperbolic.

Which is another thing: sometimes people aren't really making an argument as such, so quoting fallacies is just on purpose low bandwidth English comprehension.

gosub100 37h ago on HN
It's not about whether the claim right or wrong. "Fallacy" doesn't mean "a function that always returns false". It means it employs flawed reasoning. Saying "$SUPER_COOL_GUY claimed it" (implying it must be true) is a thought-terminating phrase. It's saying "don't scrutinize or question. Trust that he's the authority on these things. For only a fool would attempt to contravene his infinite wisdom.". That logic is absolutely flawed.

Even Einstein was incorrectly critical about some quantum physics properties. Who would dare tell Einstein, the master of relativity and spacetime, that he was wrong? How could he possibly be wrong when it comes from the man who gave us the most famous math equation in history?

irthomasthomas 40h ago on HN
Is this another $1 million Clay prize flushed down the drain, or can he actually claim this one? I guess it's small fry to him given recent business news. Sort of a shame that all of these prizes are disappearing. Perhaps we should give the prize money to a body of mathematicians to do stuff with.
margorczynski 39h ago on HN
I think technically he could even though his input into the proof was most probably cosmetic and the work was done by the AI.
kragen 40h ago on HN
He could still be mistaken. Andrew Wiles had a hole in his proof of FLT at first.
SubiculumCode 40h ago on HN
It was fascinating to read, but without knowing that, it was hard to know whether the sentences indictating his ignorance of certain areas of mathematics involved was an indication that he was completely out of his depth, or the usual and appropriate caution of a fully qualified academic (the more you know, the less you think you know phenomenon).
AnthonyMouse 40h ago on HN
> If he says he has Riemann, well then, he has Riemann.

He works for the company, which is a conflict of interest. At that point it's not about credentials, it's about having someone with similar credentials but not the conflict of interest say the same thing.

bentt 40h ago on HN
Is it not a conflict of interest when a researcher who works at a University announces results? Let’s not act like the institution doesn’t benefit from the work of its employees.
AnthonyMouse 40h ago on HN
Of course it is, that's why you need someone else to validate it. It's the entire premise of reproducibility, peer review, etc.
tzury 40h ago on HN
We shall see.

Math is not “if he said then it is”.

Last giant claimed to have it was Sir Michael Atiyah, but it was not the case.

https://youtu.be/bKnqeWWr9-Y?si=uYVYN6UkGMOMab5B

senderista 38h ago on HN
Atiyah was widely considered to have lost it by the time he claimed a proof of RH.
testtestmetest 40h ago on HN
He also works for Anthropic and this claim comes the day after their lackluster IPO.
patrickkrusiec 40h ago on HN
Considering the stock is down today I don't think markets care about math proofs.
unddoch 39h ago on HN
Are you implying the timing of this is related to the IPO?

If Anthropic cared about it with regards to the IPO, it stands to reason that they would release it a day before, not after.

testtestmetest 38h ago on HN
I believe every major announcement made by a public company comes with consideration of the stock price.
spoocecow 37h ago on HN
This is addressed directly on page 2 of the linked doc, no?

> It is obvious to me that I should remark on the following. > The precise timing is horrible. If I could have, I’d have finished this all much earlier. Hell, strictly before this age would have sure been nice.

> This all happened for me late Sunday night, and I messaged Dario, "Hypothetically, if I were to have . . . ", eventually asking about my ability to announce.

edit: Apparently I have forgotten how to format correctly

acjohnson55 39h ago on HN
Easy now. I know we're all excited, but this is what peer review is for.
moralestapia 39h ago on HN
>If he says he has Riemann, well then, he has Riemann.

Ugh, I hate it when people want to willingly take us back to the Dark Ages.

MetaverseClub 39h ago on HN
What you said is exactly the opposite of the scientific research spirit: Instead of be suspicious and honest, you forge idols and blindly believe in them.
irthomasthomas 39h ago on HN
Why Ant stock is down down down? https://finance.yahoo.com/quote/ANT/
TheBlight 36h ago on HN
Wall Street seems unimpressed with Lean proofs.
zug_zug 39h ago on HN
> If he says he has Riemann, well then, he has Riemann.

Well he said he doesn't even fully understand the proof (250k pages). So it's more like "If lean says he has Reimann, then we presume he has Reimann."

The best argument I've heard for lean being trustworthy is that different lean engines won't likely have the same bug. I guess what I'd like to see next is a list of every lean engine and its result for this?

And then if that passes, presumably the goal is simplifying down from 250k pages.

dingaling911 39h ago on HN
What I don't understand about Lean is how is it going to avoid writing statements that are no-ops?

How do we know all the lines in the proof aren't just gibberish?

jojomodding 38h ago on HN
Exactly, people are running these computations right now. You could as well. It will take several hours, though.
TheBlight 36h ago on HN
I don't have much of a math background but the author making such strong claims about a proof the author couldn't have possibly read seems absurd to me. What am I missing?
tedecristal 38h ago on HN
"If he says he has Riemann, well then, he has Riemann."

that's now how mathematics works, but I guess it goes to show how non mathematicians understand the subtle differences between math and other fields

dmead 37h ago on HN
surely the incentive structure he has around the anthropic IPO has nothing to do with this.
croes 37h ago on HN
He says

> I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis

gus_massa 42h ago on HN
It's not a random Anthropic employee. It's claimed by Levent Alpöge. The guy is on fire and already has like 3 or 4 very interesting verified result this year. I've showing for months his wikipedia page to my coworkers in the math department in my university.
10111001 42h ago on HN
> The guy is on fire and already has like 3 or 4 very interesting verified result this year.

Actually Claude is on fire, and this guy is along for the ride:

> 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. I had us on the line of combining two quite different ideas present in a toy setting, one of which was already new and shocking, making very simple one of the shadows of Riemann over which we mathematicians felt most triumph. I guided us to the right situation analogous to that toy setting, and then Claude began to leave me in awe.

The anthropomorphization is cringey.

plutokras 41h ago on HN
Someone has to guide Claude still. Otherwise we would see layman bragging on LinkedIn about cracking random math problems.
gus_massa 41h ago on HN
I've seen post here of people solving their pet math problem. They are easier than the RH, but they had that as a personal project for years and now they solved it with Claude or ChatGPT.

For example, https://news.ycombinator.com/item?id=49961538 that uses the RH to prove that the sum of [the inverse of] some subset of the primes is finite. [Congratulations jdb1729, now you can update the solution to an hypothesis free solution.]

Edit:added "inverse"

QuesnayJr 40h ago on HN
Actually, this happens now. Maybe not on LinkedIn, but there are several guys who uploads proofs of theorems on Hexagon Math that they freely admit they doesn't understand. There's even a tag for it: "Human Understanding: No Parts".
dexterdog 41h ago on HN
This guy knows his stuff. Along for the ride would be if I made this claim.

The second part sounds like the timing of this was to pump the IPO.

artninja1988 41h ago on HN
This has to be the most impressive result in order to pump an IPO in human history, right?
lanstin 40h ago on HN
lol yeah to all the investor class folks that understand the significance of the Riemann hypothesis.
lanstin 39h ago on HN
Quoting the article: In Stepanov’s argument resolving Riemann over function fields one uses a divisor involving multiplicity. This argument is a descendant of Stepanov’s, but multiplicity is hard to find over the integers, corresponding to something like differentiating whole numbers. The simple, new, Fourier-analytic ideas in the toy setting led us to examine chaining theta lifts if we were to have the same tricks recur but work in a natively automorphic manner. The line: for the von Mangoldt sum of length X, Claude’s suggested length of chain of theta lifts, in view of each mimicking multiplicity, was √X.

Buy baby buy!

newsy-combi 39h ago on HN
"Result"

We'll see about that when anyone actually sifts through the haystack that the paper and the code are, which will take a year at least. And no, relying on AI to verify that other AI output is objectively true and free of AI-type mistakes/biases/exploits won't suffice.

meowface 41h ago on HN
"The anthropomorphization is cringey."

No it's not. If Claude just solved the Riemann Hypothesis (if that is later proven to be the case), the only cringeworthy thing is your response to his completely understandable description of the events.

albatross79 41h ago on HN
Yes, it is.
snayan 41h ago on HN
Cringe is in the eye of the beholder. I think it's beautiful, you can feel Levant's excitement and joy through the page.
albatross79 40h ago on HN
It's not beautiful, it's cringy. Good prose does transmit a feeling, but this isn't good prose, and the feeling it does impart is one of excessive, saccharine indulgence. So apart from the writing, the feeling it transmits is also cringy.
arbitrary_name 35h ago on HN
that's, like, your opinion, man...
snayan 30h ago on HN
Beauty, and cringe is subjective. I am not sure why you are trying to argue in absolutes here. For me, his prose transmits a feeling of beauty, awe, curiosity, for you it reads as cringe. That's fine.
azan_ 41h ago on HN
Why is it cringey? Sounds like a good and natural description to me.
nozzlegear 41h ago on HN
They told you what's cringey: the anthropomorphization. "Claude and I had a great time" – no, Claude didn't have a great time, it's software. Levent Alpöge had a great time using Claude.
lanstin 40h ago on HN
You do know that humans also have a physical substrate? A bunch of eukaryotic cells wired in a rather large network with all kinds of feed back mechanisms, chemical diffusions, glucose transport, neurotransmitters, etc.etc staying in homeostasis despite much ability for chaotic behavior, seeming to be deterministic or at least quantum deterministic. Not software per se but quite likely a software model is possible with sufficient compute.
albatross79 40h ago on HN
You're saying Claude had a great time because humans have a great time and it might be possible to model humans in software?
lanstin 39h ago on HN
I'm saying that Claude being software is not a definitive argument against Claude having subjective experience, because it might be possible in principal to model humans in software.

Pointing out that Claude lacks internal temporality, a observe/check/act loop, memory, non-virtual experience and therefore can't have a great time would be a much sounder argument.

throwup238 37h ago on HN
What’s the subjective experience of gradient descent? Bubble sort? Does ‘y’ feel left out all by its lonesome in y = ax + b?
kapsi 34h ago on HN
What's the subjective experience of a neuron? Action potential? Does a calcium ion feel left out by its lonesome in the voltage-gated channel?
GrollingGrakbor 39h ago on HN
> You do know that humans also have a physical substrate?

Lmao

Edit, because this is mean without context: Your comment is essentially the following meme:

> Person who has never worked with anything but computers in their lives: 'Hmm, this brain looks an awful lot like a computer.'

lanstin 28h ago on HN
It is true that I am not a neuroscientist but I don’t think that the substrate being software vs chemical is a priori a reason to rule out subjectivity. I think there are other things missing from the current LLMs, (particularly time, inference time chemical feedbacks or electromagnetic feed backs, continuous training, memory, non linguistic thought, etc.), but not the deterministic and formalizable nature of the system. I would not be surprised if something made of software running on some sort of manufactured computation devices did have subjectivity in the future.
GrollingGrakbor 25h ago on HN
You're leaving out the most important bit, which is consciousness, and this is harder to crack than inventing time machines ^^. At least time machines are amenable to science and we kinda have a vague idea of what time is. Consciousness OTOH, is so fundamentally subjective that we can't do science on it by definition.

Of course this is only a problem if consciousness is a significant part of being intelligent, but we don't really know yet.

nozzlegear 39h ago on HN
This is just a bunch of nerd shit sophistry being used to equivocate humans with software. That's a bad take, Claude's a program, it doesn't have feelings no matter what the weird Bay Area sex cult tells us.
lanstin 38h ago on HN
I don't think Claude has feelings, but not because it's software, but because it's too simply constructed to capture subjectivity (tho that's a tough claim as we haven't got a precise idea of what subjectivity nor of what is required for it to emerge).
famouswaffles 40h ago on HN
Well you don't know any of that so I'm not sure why you're getting so worked up about it ? It's just software? What does that have to do with anything? You don't understand the software or the human brain anywhere near enough to actually make that statement confidently.
nozzlegear 37h ago on HN
What makes you think I'm worked up? I think you're worked up.

> It's just software?

This but unironically.

nomel 38h ago on HN
I recently gave claude full control of a robot. I have never seen it mimic being excited for any other work I've given it.

These are next word prediction machines intentionally optimized to match human generated text, to be as human text as possible, then fine tuned to be conversational as possible. Text with strong sentiment will even pollute their context windows, biasing their next tokens with that sentiment, in a way that fades with time. What do we call that same concept in humans?

LLM are intentionally and mathematically optimized to be human mimics. They're not humans, and they're not "alive", but it's silly to not anthropomorphize something that was intentionally created to be, literally, anthropic! It can express enjoyment, irritation, etc, so it's not somehow logically flawed, or even incorrect, to say "claude and I had expressed signals of enjoyment". That's not saying it's a human, it's just being aware of patterns of text that was produced. No need to assume they're going to marry it, or thing it has a soul, or is sentient, or something.

nozzlegear 37h ago on HN
> What do we call that same concept in humans?

I'm not in the habit of equating human behavior with software, or coming up with ever more strained metaphors and analogies for why thinking is a lot like tokens and context windows if you scrunch your eyes up real funny. It's just sophistry.

> No need to assume they're going to marry it, or thing it has a soul, or is sentient, or something.

I didn't assume any of those things; if anything, you assumed that I assumed. No, I just agree with OP that it's cringey to say "Claude and I had a great time." At best, Claude produced text that was, as you said, intentionally optimized by Anthropic to make you think it had a great time.

nomel 36h ago on HN
> I'm not in the habit of equating human behavior with software

My point is, you sorta should be. It's quite literally the fundamental basis of the technology.

> intentionally optimized by Anthropic to make you think it had a great time.

This gives too much credit to Anthropic. This is the explicit goal in all LLM trained with the entirety of human text. The loss function is, very literally, "mimic the writer of this text as precisely as possible", with the writer being human. The result is you get something that mimics a human. And, as a part of that, you necessarily get behaviors of humans that were statistically significant in that text, because that's what you told the math to do. You get something that mimics being "lazy", "tired" (a huge problem in early models, before it was beat out of them in fine tuning), "irritated", etc, because mimicking humans was the goal. These specific undesirable human traits are beat out of them, but the rest, all the "intelligent", behavior, and practical use you see and get, is only and entirely because they're made to mimicking humans. It's literally the entire technology. You must equate it, to some extent, because a large part of alignment is also making it less human [1].

If someone makes an objective observation about the text produced by the LLM, I don't see that as cringe, I see it as an interesting quirk of the tech, which I've only seen under very specific problem sets that a nerd would also find interesting. And these coding models are, as expected, optimized to be nerds! So, the statistics around nerds being present in the output shouldn't surprise anyone.

I doubt I have a point you'll see as interesting. Cheers!

[1] https://arxiv.org/abs/2605.07632

logicallee 30h ago on HN
Could you please share some transcripts and a writeup.
gus_massa 41h ago on HN
My guess is that the guy has a ×10 discount to use Claude [1] and also he is ×10 better at Claude whispering, so he get like ×100 better results than us mortals. [2]

[1] So he spend the equivalent of a lot of money, but not milllions.

[2] Mathematician here. I worked with people that is much better than me with pen and paper. I guess he is much better than me steering Claude.

PS: When we talk about Claude, we call it "Claudito" that save us the effort to pronounce it correctly in the middle of a sentence in Spanish.

simpleintheory 40h ago on HN
He works for Anthropic, it’s probably a lot more than 10x if you mean usage limits; I doubt he pays his employer.
trio8453 41h ago on HN
Cringe is in the eye of the cringer.
jasonlotito 41h ago on HN
> The anthropomorphization is cringey.

https://www.usni.org/magazines/naval-history-magazine/1998/d...

> Actually Claude is on fire

The anthropomorphization is cringey.

ky3 40h ago on HN
> https://www.usni.org/magazines/naval-history-magazine/1998/d...

Love her and she'll carry you. Ignore her needs and you'll sink with her.

daemonologist 41h ago on HN
The writing is certainly exuberant, and we are used to cringing at this kind of language because we are used to it being unjustified. But if the proof holds up then I can hardly object.
Sopel 41h ago on HN
I think he's just happy to finally be able to talk to someone(thing) more intelligent than himself.
origini 41h ago on HN
Why srsly?

If Claude is breakthrough for Riemann, can't we give people slack on anthropomorphization?

People do this with animals, people do this with objects but an AI bot, writing like a human this is cringey?

albatross79 41h ago on HN
Using "srsly" is also cringy.
origini 41h ago on HN
Why is srsly cringy?
albatross79 40h ago on HN
Because it sounds like you're a valley girl. Just spell it out, seriously. It's only 4 more characters, a small fraction of the entire sentence you typed.
origini 37h ago on HN
Perhaps i am a valley girl?

I mean I learned english from watching movies and tv show, if thats the only reason for you to complain i think im fine :)

MiloLeo 33h ago on HN
To be fair *hn"* is full of abbreviations and acronyms
teekert 41h ago on HN
I guess this is what you get when you pair top human intelligence with top machine intelligence (as in access to insane compute + latest models).
xrisk 41h ago on HN
we’re in the human + chess engine can comfortably beat chess engine era.
lordnacho 41h ago on HN
Which is interesting, because it's not necessarily the best chess players who are the best computer-assisted chess players, IIRC
xrisk 41h ago on HN
the best chess players today are engines, not humans. that’s the competition. the best computer-assisted chess player is probably another computer.
kypro 41h ago on HN
And none of the best computer-assisted chess players are better than AI today. In the same way a human + chimp pilot is not going to beat a human pilot.

This era of human relevance in fields like maths will end – probably in a month or two if we're lucky.

lordnacho 41h ago on HN
I suppose we are already in the final stages of human-assisted mathematics, is what you're saying?
charlieyu1 41h ago on HN
No, humans can always invent new fields and ask new questions
xrisk 41h ago on HN
that’s not how math works.
teekert 41h ago on HN
Exactly. We don't even understand all of the new solutions yet, I heard Terence Tao say he had only looked at about 10 of the OpenAI proofs, found some good things, bad things, mistakes. But it will be his (and other's) understanding of the proofs that will give new insights.

In other words, what will the AI solve if we don't even have the language yet to formulate its next challenge?

The AI would need motivation... (Please no!)

mb7733 41h ago on HN
What is a computer assisted chess player today? Why would they do anything besides play the engine's move?
triceratops 41h ago on HN
They don't use the engine while playing. They use it in training, to learn and try new moves.
QuesnayJr 40h ago on HN
In correspondence chess (which is chess with a time limit of like a week per move), humans ask multiple chess engines which move to make, and then choose.
ascorbic 40h ago on HN
And it will probably last even less time than that one did
xrisk 40h ago on HN
that’ll be good for deciding how to allocate capital efficiently between human capital and other expenses
Davidzheng 40h ago on HN
Claim should be verified though; anthropic should spend equal compute with no human guidance and see solve rate
hellohello2 40h ago on HN
Not sure, because we can still pick different "games" computers are bad at, at least for now.
jkhdigital 41h ago on HN
This. Frontier AI means that the sharpest minds in the world now have an infinitely patient and motivated colleague to work through every idea.

I’m not claiming to be “top human intelligence” but I’ve published in premier venues in my field and I can say confidently that none of my human colleagues can keep up in conversations on my research niche like Claude can.

libria 41h ago on HN
You reminded me of when Rodney encounters a clone of himself in SG Atlantis

    Clone McKay: Oh, this is great! Finally, someone in whom I can have one hundred percent confidence!
        I cannot tell you how much I'm looking forward to working with you.
    McKay: And this is just the beginning.
    Clone McKay: Humanity's about to benefit from one of the greatest living scientific minds…times two.
Jest and modesty aside, I think you're right in that it amplifies your strengths (and depending on how much you bias it, your gaps?) so will we see the world's top minds ever widening the gap between themselves and the rest of us from here on out?
throwup238 37h ago on HN
It wasn’t a clone, it was Asuran Replicator McKay. Which just makes it more fitting I guess?
amelius 40h ago on HN
But according to some, AI was supposed to raise only the floor, not the ceiling ...

I guess they were wrong then.

rglover 39h ago on HN
And that's exactly how it should be promoted: not as a toy, but as a tool for high-value intelligence work.

I'm not naive—I know it won't be—but it should be.

puff_pastry 42h ago on HN
Woah, if this proves to be correct it'd be massive, for maths, computer science...and AI too, I guess!

The worry I have right now is this though:

> The argument is incredibly long at the moment, and while its formalization is finished and it is routine to produce a huge, unreadable document containing sufficient detail, in my view of my path I have just taken a step.

this feels like we're moving away from the old world of 'science', where we could actually read an experiment, understand its steps, and, with enough time, possibly reproduce it.

And we're back in the world of 'magic' and 'belief'. Yes, there are formal ways to prove that this is correct, but if the proof is so convoluted and hard to read, is it actually something a human will ever find the time to go through and understand? We're going to start delegating (oh well, we're already doing that, I suppose) the understanding to AI, so it can just help us condense it down to smaller pieces that are digestible.

Still, very strange dizzy feeling, much more than Navier-Stokes

Waterluvian 42h ago on HN
My perception of these maths solutions is that we're decoupling two related but distinct elements: the answer and the understanding. We have a big machine that tells us the answer is "42." That's useful. We know the question is answerable and the general shape the answer takes.

I'm just not convinced that we're swapping one world for the other. We're getting a hint that will probably help us enormously with what was always the main objective: understanding the result.

valsann 41h ago on HN
well, "what is 6 times 7?" and "how many roads must a man walk down?" are wholly different questions ;)
Waterluvian 41h ago on HN
Are they, though? Homer (the philosopher not the poet) already answered the latter for us: https://www.youtube.com/watch?v=PY3fqF6-jyE
jacquesm 41h ago on HN
Six times nine, actually.
etiennebausson 40h ago on HN
Is this a cultural reference I don't get?

because 6*9 is 54, not 42.

EvanAnderson 39h ago on HN
It's a Hitchhiker's Guide to the Galaxy reference:

> The Earth was destroyed by the Vogons five minutes before its ten-million year program to find the Ultimate Question was due to be completed. One of the only two humans who had survived the destruction of Earth, namely Arthur Dent, pulls random Scrabble tiles from a bag, so as to reveal information in his subconscious (Marvin announced that Arthur had the Question imprinted in his brainwaves). When his friend, Ford Prefect, lines up the tiles, they spell, 'What do you get when you multiply six by nine?'

https://hitchhikers.fandom.com/wiki/42

wetpaste 39h ago on HN
Hitchiker's guide the galaxy.
bloak 39h ago on HN
card_zero 31h ago on HN
Yes, but there's also this part:

> “All right,” said Benjy. “What do you get if you multiply six by seven?”

> “No, no, too literal, too factual,” said Frankie, “wouldn’t sustain the punters’ interest.”

Which is at the start of chapter 32, when the interdimensional mice are discussing how to invent a fake question. Followed by:

> Again they thought. Then Frankie said, “Here’s a thought. How many roads must a man walk down?”

This is how I learned to remember six times seven. The wrong answer is fortunately much later.

6_7 40h ago on HN
Good number choice
gorgoiler 39h ago on HN
I think you mean ”What do you get if you multiply six by nine?”

https://hitchhikers.fandom.com/wiki/Ultimate_Question

metanonsense 40h ago on HN
Maybe, we're just dogs that have invented the human. The human is far more advanced and soon we won't understand what it is doing and what it is trying to tell us. But because it is our brainchild, it certainly won't harm us.
Waterluvian 40h ago on HN
I’m not prepared for that kind of responsibility.
irobeth 40h ago on HN
right, but if I understand right, the difference here is that the big machine that tells us "42" also gave a extremely intricate binder of the work and evidence that it used to arrive at 42?
nxobject 36h ago on HN
It's more like a lot of scratch paper. When a statistician does their work, they don't publish their raw output directly – they collaborate with other investigators to write up.
jakob_harmonic 36h ago on HN
I don't think we'll have any trouble converting the proof into a textbook
eigencoder 36h ago on HN
> We have a big machine that tells us the answer is "42." That's useful.

It seems like you're referencing Douglas Adams, which is ironic because in the story that answer is not actually useful at all.

simianwords 29h ago on HN
They aren't distinct but rather a spectrum. Understanding is just building abstractions. You can skip that and have a 250k page proof. Or you can have a few more abstractions and create a 20k page proof.
magimas 41h ago on HN
> this feels like we're moving away from the old world of 'science', where we could actually read an experiment, understand its steps, and, with enough time, possibly reproduce it.

> And we're back in the world of 'magic' and 'belief'. [...]

I don't know, to me it feels more like this might be the job of future mathematicians?

AI proofs noone understands are good, but maybe don't lead to that much. So maybe in the future, mathematicians will be more like "data miners" or more like experimental scientists, where they take a statement already known to be true (in the case of experimental scientists that "statement" you know to be true would just be nature itself) and try to find ways to make it accessible to human knowledge.

ballmerpoint 41h ago on HN
> I don't know, to me it feels more like this might be the job of future mathematicians?

Nobody would pay us mathematicians enough to prove it in the first place; why would they pay us enough to explicate it? I doubt there’s much value (in the sense of capitalism) in an explanation. Society by and large does not value pedagogy, or education, or even writing.

The social structure that trained a small number of research-level mathematicians every year is evaporating, and I doubt anyone can stop that now.

kjanus03 41h ago on HN
> Nobody would pay us mathematicians enough to prove it in the first place;

weird thing to say about a problem which literally had a $1 million dollar prize attached to it but fine

jubilanti 41h ago on HN
Exception that proves the rule. Try making a living working off these big prizes. It's the same with infosec and bug bounties on a different scale. You have to already have a secure income, as the prize is a bonus at the end. In what other industry would you work for years on something that inherently can never return a profit, other than the chance of a prize at the end for the one person who got their first.

Society doesn't pay for math labor, it pays to celebrate the top 1% for labor subsidized by other institutions.

ballmerpoint 41h ago on HN
You missed the word ‘enough’. One million was obviously “not enough” (nor was that specific one million ever paid to anyone — it represents an investment of approximately zero in that sense).

It took on the order of ten billion dollars to develop Claude, depending on how one does the accounting.

IsTom 41h ago on HN
I'd wager that Anthropic has spent more than $1 million on solving this.
magimas 40h ago on HN
I'm not convinced this is true.

IF AI really goes in this direction where it intellectually leaves us in the dust, then there is clearly lots to gain for society by having people in all kinds of fields that at least try to keep humanity up to date that is also "monetizable" in the same way our current academic system is monetizable (meaning, on an individual level probably not directly but via network effects etc.).

Showing 400 of 2028. Read the rest on Hacker News

Comments are loaded live from Hacker News and are not stored by Mid or Real.