logoalt Hacker News

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

1832 points • by iamronaldo • today at 2:16 PM • 1248 comments • view on HN

Comments

utopcell • today at 4:03 PM

Thank god for Grigori Perelman, who isolated himself for a decade to prove the Poincaré Conjecture, and we can at least claim that one of the millennium problems was not solved by a machine.

fourseventy • today at 3:08 PM

we were 2 days shy of being able to say "we got reimann hypothesis before GTA6"

u1hcw9nx • today at 6:50 PM

Hurried Announcement: Alessandro Masullo GRH https://math216.wordpress.com/2026/11/20/hurried-announcemen...

Is this yet another case of AI company stealing almost ready proof form someone else?

Dropoutjeep • today at 5:08 PM

> 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.

I truthfully don't know enough to comment on the validity of a 1.3Gb lean proof for the Generalized Riemann Hypothesis, but is there any way to read this paragraph other than fluffing Claude/Anthropic right after IPO?

➕ show 2 replies
Kuyawa • today at 5:18 PM

"Riemann hypothesis is considered the most important unsolved problem in pure mathematics and a Millennium Prize Problem"

So what's next? If it can solve the hardest I guess all problems are a piece of cake. Does it create its own problems to solve now? Does it wait for us to create more problems? If it creates its own problems and solve them with clear benefit to humanity, then we're entering a new Age of Enlightenment but this time we're just spectators

11/15/26 GRH (it says 11/16 on the pdf but most of the final work and eureka moment were on the 11/15 Iterated theta lifts...)

➕ show 1 reply
chubot • today at 5:25 PM

Funny: Claude didn't know that was claimed to be proven by Anthropic today, even though it said it searched the web.

Prompt:

what's the big deal about the riemann hypothesis? is it easy to state?

Claude:

(explanation)

Recent news: it’s still unproven. Last month there was a notable related result. OpenAI claimed to settle the “quasi-Riemann hypothesis,” a weaker version of the Riemann hypothesis, which is arguably the biggest open question in math. Roughly, the weaker version says the zeros stay inside some strip around the critical line, not exactly on it.

albinstman • today at 2:51 PM

Interesting timing, given that the claim comes just a day after Anthropic’s IPO.

➕ show 3 replies
m4rtink • today at 5:22 PM

Oh, the "stonks-please-go-up!" caned news triggered by ANT going below $90!

Cool, I wonder what they have prepared for $80 and below - feels like an advent calendar, keep going! :)

swingboy • today at 5:47 PM

Does the published content have his prompts? I’m assuming a world-class mathematician with unlimited usage guiding Claude along gives a bit more advantage than an average Joe prompting “pls solve the Riemann Hypothesis”. Would the same conversation in Astra produce the same result eventually (stochastic variation aside)?

Painsawman123 • today at 5:27 PM

It's funny how people are using this supposedly successful proof of the RH by a swarm of AI agents with the aid of a human, as evidence that "AGI" or advanced intelligence is going to emerge within a few months... When in reality , solving the RH means very little in the grand scheme of things! When will people wake up to the fact that the progress we've been seeing over the past few months has been exclusively(!) in "verifiable" domains and that no(!) progress has been made in areas where the signal_to-noise ratio is low and that Furthermore this "verifiable domain" approach isn't a general form of intelligence?

If humans are to build "Agi" or superhuman intelligence or whatever in the coming months/years, it'll be through a completely different architecture than autoregressive models a good measure of progress would be to see a swarm of Aa agents solving something like "the continual learning problem" or coming up with new architectures that are more efficient. This is how progress in Ai should be measured not by how many mathematical problems are solved (which are impressive in themselves but tell us very little about progress toward AI that can handle the noisy real domains)

prmph • today at 3:33 PM

The writing is kind of weird, was it produced by Claude as well?

> 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. 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.

Tens of such occurred...

pcj-github • today at 4:55 PM

What struck me is the hubris and self-grandeur that comes through in the PDF... "I", "I", "I", ..., he clearly has a great emotional connection it...

To what credit does this person deserve in the outcome? Should this person be remembered by history? For the $1,000,000 prize, how much should he get? He self-admittedly does not even understand how it works.

jrflo • today at 2:38 PM

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.

➕ show 1 reply
ozgrakkurt • today at 6:22 PM

Wait, wait one minute, don't sell the shares please

rf15 • today at 3:10 PM

> The argument is incredibly long at the moment

so... we have plenty of places to hide logical flaws, got it

➕ show 1 reply
enahs-sf • today at 3:41 PM

I wonder if the history books will look at famous problems solved with AI as the Barry Bonds asterisk of proofs. Euler, Gauss, Reimann were natty but we now have access to intelligence enhancing drugs.

dash2 • today at 4:29 PM

Just reading about this, and my three-year-old has just shown my two-year-old how to complete his jigsaw puzzle (which has three pieces).

Perspective.

num42 • today at 3:22 PM

If this is true, it would be huge. OpenAI would race desperately to tackle the remaining Millennium Prize Problems. Perhaps they would even turn their attention to solving P vs. NP with unprecedented urgency.

lynndotpy • today at 3:22 PM

Going to be hard getting used to saying "the Riemann Theory".

Amekedl • today at 5:57 PM

inb4 its just claude convincing even more humans (trust AI managed to convince me at gpt3.5)

MiiMe19 • today at 6:44 PM

Well, now that we are at this point I have to decide if I should kill myself or AI execs. Maybe I'll ask ChatGPT.

amelius • today at 5:10 PM

But who can claim the millennium prize now?

BrokenCogs • today at 4:38 PM

This is crazy, I wonder if they sat on this until the IPO?

➕ show 1 reply
LandoCalrissian • today at 4:21 PM

Bet you he didn't.

jacquesm • today at 3:44 PM

Did mathematicians not working for Anthropic have access to this model as well?

ChaseRensberger • today at 5:45 PM

man, i guess we didnt get a solved reimann hypothesis before gta 6

lionkor • today at 3:48 PM

As a complete amateur; what's the practical difference between knowing that there's a proof but not understanding it, and not having a proof?

Is it useful to know that there is a proof, in the unironic "trust me bro" way, such that we can rely on it to be proven yet we don't understand the proof? Or is having a (currently, for humans) unparseable proof just as "practically useless" as not having one but believing one could exist?

➕ show 1 reply
jsrozner • today at 6:08 PM

I'm certain the guy working on this is much much better at math than I'll ever be. But his remark.pdf looks like it was written by someone who has been spending too much time with Claude.

We used to call this style of writing delusional. Now we call it a proof of RH.

sirnicolaz • today at 4:12 PM

I am somewhat happy to finally a title mentioning a human solving the problem and not a generic "AI". There is always a person driving the model towards the solution, this shall never be forgotten.

empath75 • today at 3:25 PM

We almost got a proof of the Riemann Theorem before GTA VI

chvid • today at 4:22 PM

175190 files totalling 17374394 lines of code.

CraftingLinks • today at 4:23 PM

24 hours later:

'Oops I made mistake and got carried away'

Wanna bet?

1970-01-01 • today at 3:44 PM

So the next problem is a map of all the primes, right?

xyzsparetimexyz • today at 5:05 PM

I'm dreading a scenario where we solve this and P=NP and figure out non causality violating FTL and all this wonderful theoretical stuff but none of it matters and the world burns anyway

yoaviram • today at 4:01 PM

Disproving it would have been so much cooler.

8s2ngy • today at 3:47 PM

This was unimaginable even a year ago. We're living through a time that will be etched in history. From my work in software development and from conversations with friends in academia, I have understood that the old order, centered on human intellectual labor, is dead: so dead that we'll never return to how things were, even a year ago. AI is here to stay, and it will keep getting better and better, recursively. I shudder to think about what will happen to humanity as AI makes the jobs of millions of people obsolete. The nerd in me wants this technology to keep developing so we can push the limits of what humanity can achieve. But I'm not naive; I know this will come at the cost of millions of people losing their livelihoods.

nickalaso • today at 4:12 PM

So, I am convinced that we are already screwed (or saved) whatever you think that statistical likelihood of AI based 'misalignment' induced human annihilation risk is.

Anyone who hasn't played this game should: https://www.decisionproblem.com/paperclips/

Its very precient.

Divide and conquer and problem/sub-problem composability can be used to solve any known problem that is a solvable one.

I understand the copium of the anti-ai guys. Its called being in denial. But we have to face reality here. Yeah, diminishing returns, yada yada. Problem is the Billionare with the most available compute can now get the swarm to push back each of these diminishing return walls simultaneously by this own swarm.

Classic attention based transformer architecture is very computationally inefficient, its O(n^2) on the input. So what happens when the ai swarm finds a universal and provabale way to make the same model with O(log(N)) or O(N) complexity? All available compute becomes exponentially more powerful overnight.

What if such a thing is impossible or too difficult? Well maybe it can just fundamentally increase the production capacity of every factory producing compute x1000 or x10000 etc. There is no power limitations for those numbers or even resource limitations when we have nuclear power, geo thermal, and the entire earth's crust at its disposal.

Ok, barring that, now it can work on improving robotics and toolsets tied to operating factories and begin exponentially building more factories until we simply multiply the number of factories producing chips 1000, 10k, etc.

Or most realistically, all of the above, always, all at the same time.

More efficient software, more efficient chips, more chips produced, more factories producing chips, etc.

Every step towards increasing the number of simultaneous coordinating swarm members will push the intelligence higher, and the capability of solving or improving its own intelligence also higher.

At this point we have at most a few years.

Hmorgan12 • today at 3:19 PM

Wow this can take years to verify, maybe even decades

➕ show 1 reply
nathell • today at 3:40 PM

17M+ lines, 1 GB of Lean in 175K+ files. Good luck of understanding all of that in one’s lifetime.

lukeplato • today at 4:33 PM

Looks legitimate to me. This is a heroic-looking ascent, likely not feasible through human-effort alone.

1. 10/6/26: Stepanov determinant. 2. 10/9/26: Dp ≡ 0 on a CM curve for p inert in K, D the invariant derivation, and Da·θ(0) Ωa is divisible by p a p + a p3 +···−O(log a). 3. 10/19/26+: qN = M k + R, using x1 · x2 + · · · + xk−1 · xk, wanting "automorphy" of N since μ is automorphic from θ on sixfold cover of GL5 as a period integral of thetas, quadric has natural interpretation on corresponding CM e.g. C/Z[ω] as Weil5, put Poincaré bundle on the edges to get θs. 4. 10/6/26+: W (1S ) ≥ 0 via a determinant, interpret as Stepanov with multiplicity one again; bootstrap!! 5. 10/20/26: definitively switch to Λ aka∑ n≤X Λ(n) ·( X n − 1) + − X log X + (1 + γ) · X; Guinand-Weil had the sign, variance test function trick to make things look trace-formula-like (but how can I pigeonhole / Turán over n? X? it?), Mehta-like integral over twists or at infinity. . . 6. 10/8/26: W( 1(0,1] + (∼ Xη ) · 1(0,1/X]) ≥ 0 gives > 1 − η. 7. 10/14/26: Determinants for W (·)’s containing the desired periods, indeed GGP periods, with values 0, 1 meaning interlacing weights, or up to m non-interlacing ones. 8. 11/15/26: Iterated theta lifts U (X0) → U (X1) → · · · → U (Xm) between totally definite unitary groups, π ↦ → π ⊞ μj [t], wins multiplicity m. Time to take m like √X. . . 9. 11/15/26: Norms of the lifts and level and cancellation of order-m terms with proven divisibility via Chowla-Selberg, Weyl, and Siegel mass. 10. 11/15/26: Sufficient lower bound on divisibility of the m-fold iterated lift. 11. 10/6/26: Iwasawa tower via analogy to pigeonhole on Fqn /Fq . 12. 11/15/26: Seesaw: the period of the iterated lift is a Bessel period independent of the Hecke characters, Rallis. 13. 10/6/26: Bost slope inequality. 14. 11/15/26: precise computation of the "cost" of each lift, showing that one would stop at 2/3 a priori, attempt to get a main term and transfer constant to a function field or kill via twisting, need for "division". 15. 11/15/26: Glue together an integral structure across Hecke characters for a "gcd" i.e. extra averaging. 16. 10/14/26: Relative trace formula, seen in a different half-plane argument via sixfold cover of GL5 automorphy of μ and reducing via Cauchy- Schwartz to a bound on the norm of a suitably projected Poincaré series, approximate orthogonality of the columns of the integral structure. 17. 11/15/26: for the classes in H1( oE[ 1 ℓ] , Qℓ/Zℓ(−e)) that we couldn’t kill by twisting, fight for a significant weakening of Vandiver.6 18. 11/15/26: computation of covolumes, theta series’ index, congruences between lifts and forms. 19. 10/19/26: precise choice of the Iwasawa tower, specifically using Behrend!!7 20. 11/16/26: GRH

qwertyforce • today at 4:19 PM

are we in the singularity now?

igravious • today at 5:54 PM

A new era for mankind has undoubtedly dawned.

All that is solid melts into air, all that is ineffable crystallizes.

Routine8502 • today at 2:44 PM

Big if true. Looking forward to seeing it verified!

chaosconst • today at 5:41 PM

I am trying to reproduce the lean on a 192-core aws machine, live on: https://cloud.infero.net/rh-en.

pfdietz • today at 3:40 PM

Says the proof has been formalized.

TaupeRanger • today at 6:15 PM

Very cool. Now when do we get the breakthroughs in real science and medicine?

sp-ti-en-ma-in • today at 3:28 PM

Is this last normal year?

🔗 View 45 more comments