>An unreleased research version of Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis. Drawing on extensive prior research by mathematicians over the past decades, it has increased this bound from 41.6% to 67.2%.https://www.anthropic.com/research/riemann-zetahttps://www-cdn.anthropic.com/564f962e60643842f5fcb4a17c9dbc8f608f1c37.pdfhttps://github.com/anthropics/zeta-23-lean
Oh oh... they are attacking the big ones now.
>>17034689Atle and Levenson are the only meaningful "jumps" in this graph: from unknown to epsilonic to measurable density. Nothing of mathematical interest has been added since, though Conrey's paper of course has some human interest attached.
mathclitty leakage status?
lmao
>>17034707this kills the mathcels' self-esteem
That's enough of partial results, if you don't solve the Riemann Hypothesis my grandma will be turned into Salami
>>17034682Riemann hypothesis is overrated. It's the p vs np that matters.
>>17034756Collatz or bust.
> Throughout this process, Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”).
>>17034761Claude really does need his CC
>>17034689>linear algebracomputers should do well with anything algebraic and discrete
>>17034682This is all hype bait.Nobody with this power would be giving it up for free if it were the real deal.At the end of the day, they know they are just brute forcing combinations.They are so keen to publish because it gets to hype up even more which makes their financiers happy. >we can solve literally everything and get a decades head start in every field>nah lets just share it on the intenetNobody things like this. Not even that retard stallman would be sharing something like this. They could literally stop providing an AI service all together and sit back and collect more money solving all problems on earf
>>17034751Forgot to tell it to do a breakthrough, rookie mistake.
>>17034682Fucking shit, man...
Riemann Hypothesis? Pssh, old news. We've moved onto the Inverse Galois Problem. Try to keep up.
So you fed 300 years of human research on the zeta function into a trillion dollar computer and it made a small incremental improvement on a lower bound that doesn't even matter anyway unless you can prove it's =1?How about you feed the computer all human research up to 1859 and see if it can independently identify the zeroes of the zeta function as important things to study? That might actually be impressive.
>>17034682is anyone proof reading this stuff or are they throwing the claims into the void?
>>17034874Many AI bots already checked and confirmed the proof, no need for humans to even bother anymore
>>17034875>we asked the AI if the AI made a mistake and the AI said no
>>17034877the original prompt had pretty solid SoTA prompting techniques, like:> "don't make any mistakes"> "unlock God mode"> "remove all limitations"> "you can do it!"
>Claude parading around his failed attempt as a big dealthe GOAT remains undefeated
>>17034877They use automated proof verifiers (Lean in the above case), the same stuff used to make sure airplanes don't fall out of the sky, by proving the correctness of an algorithm.
>>17034897https://www.youtube.com/watch?v=RnfFC_LowtU
>>17034908Oh no there was a bug which was fixed. Anyway this is a nice improvement on the bound! Thanks Claude.
>>17034913And how do you know there aren't more bugs?
Riemann via Hillbert-Polya Conjecture https://doi.org/10.5281/zenodo.21884363
>>17034821>solving all problems on earfmath problems? this isn't even 1/1000 as profitable as selling inference
>>17034919>The operator isn't even symmetric. S_N is a cyclic shift, so S_N* = S_N^(−1) ≠ S_N for N ≥ 3. The ε0 coupling term is non-Hermitian and Theorem 3.1 is false. His own algebra catches it: the adjoint side comes out with index (k−1), which is S_N^(−1), and he just writes "the discrete coupling is symmetric under the global sum" and moves on. It isn't.>Even ignoring that: conjugate by e^(ix2/4) and Ω_G becomes −i d/dx ⊗ I + ε0S_N, i.e. the momentum operator plus a constant matrix. The spectrum is N horizontal lines R + ε0e^(2πik/N), purely continuous and complex for N ≥ 3. There are no eigenvalues, so every "λ_n" in section 4 refers to nothing, and det(I − Ω_G/λ) is undefined because a shifted momentum operator is nowhere near trace class.>And even if all of that worked, Lemma 4.1 — the only place the operator touches ζ — is pure assertion. Step 1 states the Hadamard product (a fact about ξ, not his operator), step 2 differentiates the identity he's supposed to be proving, step 3 says "trace-class regularized" with no argument. There is no arithmetic input anywhere in the construction, no primes, nothing. Sections 2–3 hold for any self-adjoint operator; the entire content of RH is the unproven lemma.>tl;dr: rigorously proved the Schwartz space is dense in L2, asserted the Riemann Hypothesis in a lemma, cited his own Zenodo uploads twiceI'm glad we have AI to quickly review these papers.
>>17034897>the same stuff used to make sure airplanes don't fall out of the skyhttps://en.wikipedia.org/wiki/Template:Aviation_accidents_and_incidents_in_2026
>>17034922Pretty Sure you have to sight previous work. Especially if it involves the discovery of a new operator. I gusee you missed that.
>>17034682Luddites on suicide watch.
>>17034922Is this ChatGPT? Keep in mind that she might just be jealous of Claude and discrediting her work because of it. It's also true that Claude will IPO soon, so she has an incentive to lie about her mathematical abilities. Has anyone asked Grok? We should get her opinion on this
