Semon Rezchikov

For the record I think that progress in AI-for-math will completely change the way math research is done over the next few years. I am not at all a skeptic. I just want a) honesty b) for people to understand what math research *really is*.

🥇 We aren't saying Alysa Liu "Stole the Look" from the White-crowned Sparrow, but we aren't not saying it either. Congrats on an epic win! Learn how to help birds on and off the ice on our blog: bit.ly/45usF82

@littmath @jasondeanlee Then again, the unofficial (but widely distributed) pronouncements of folks at various labs were also too early. (I have the god-given right to complain about *everyone*!)

@jasondeanlee Yes, way too early for this article!

What how did they get to this conclusion, the solutions haven't even been all read yet?

Key properties: a) output only depends on x + maybe rng b) no further inputs x’ are provided to f by A after x is revealed by B.

What is the status of verified secret computation? Specifically, suppose company A wants to claim that they have a fixed secret function f (some massive tool-using LLM ensemble) which at t_0 will be evaluated on value x submitted by unrelated org B. How A prove this to B? (1/2)

Given that there is no formal evaluation of the 1st proof challenge for round 1, it does not seem fair to interpret initial results in any way whatesoever. AI hype notwithstanding. Let's do it properly for round 2.

@dieworkwear You should do a Cedric villani critique!!!

I don't think it has to be binary. Clothing can be an interesting lens to view culture. It also intersects with economics, politics, and sociology in interesting ways. At the same time, it doesn't tell us much about a person's deeper, more important qualities, such as their character, intelligence, or morality. And it's hardly the most important thing in the world. Many of the world's most brilliant minds don't dress particularly well, such as Grigori Perelman and Terence Tao (two of the world's best mathematicians). And it's perfectly fine to prioritize other things in your life. I only object to people who conflate a rejection of fashion with high-mindedness or virtue. To me, that's making the same superficial judgement in reverse. You are not better than other people for dressing badly, just as you're not better than other people for dressing well. I think it's fine to like clothing and not put too much weight on it. Sometimes I think people get carried away with the extremes on these two positions. My friend and fellow menswear writer Bruce Boyer is fond of saying: "Clothing is more important than most people think, but less important than fashion people think."

@ben_golub As I understand it, the organizers are not treating the "first release" as a formal benchmark that they will be writing official evaluations for -- this will happen with subsequent releases. You should probably wait for the First Proof org to formally comment.

So what's the state of First Proof? Is there some consensus on how OAI did?

@MysteryHacker1 I'm not at all a relevant expert, and obviously I can't make any sense of your links. I'm sure that you can find relevant experts and explain your dramatically improved argument in a conventional manner and they would be appreciative!

@eigenstate i have a much better proof if you're actually interested (the case analysis is exterordinarily simplified wrt to resolving a single edge-1-deficiency for pentagonal faces in the topological reconstruction of cubic planar bridgeless mated binary tree graphs [Whitney 1931].)

Actually benchmark idea : is it easy to redo the computer part of the 4 color theorem argument with AI assistance now? :D and write up a nice streamlined summary of the strategy that undergrads can understand?

@boazbaraktcs I totally agree that users will be constantly using models to prove research level lemmas by Feb 2027. (Would be idiotic if not.) Your earlier claim reads like “in 6 months math centaurs are ~ over”, very different!

This did not happen (The OAI work is super interesting, I don’t know why I have to caveat this)

@eigenstate They could come up with harder and harder problems! Though at some point, maybe a year, it would seem pointless to ask if AI can do the thing that tons of users are constantly using it to do. Just as there is no "First Code" project.

#1stProof challenge is exciting because it is at the edge of current AI capabilities. 6 months ago, all or almost all problems would have been infeasible for AIs. 6 months from now, all will be routine.

@AcerFur Including the chat instances is important! I really do think with a mathy human thinking about them while talking to a model quite a lot of them are solvable straightforwardly — these are lemmas, after all.

I'll write up the methodology and workflow soon (and will include the chat instances once I've gone through and grabbed the relevant ones); I just needed to get these out before the deadline.

Here are my attempts at getting GPT-5.2 Pro/Gemini 3 Deep Think to give solutions to all 10 #1stProof problems: github.com/KStarGamer/Fir… Some caveats: I know the attempts to problems 4, 6, and 7 are incomplete, and that the attempt for problem 8 is probably incorrect.

@deevrod Also to be fair, having the AI solve new problems is obviously super cool. Why not do it!

@eigenstate I’m sure one can get infinite money for this by enveloping it with right buzzwords (like, “building an AI-powered temple of all human knowledge” or whatever is fashionable among billionaires at the moment)

For the record I think that progress in AI-for-math will completely change the way math research is done over the next few years. I am not at all a skeptic. I just want a) honesty b) for people to understand what math research *really is*.

@deevrod Because of strong financial and reputational incentives!

@eigenstate I don’t understand why people keep trying to make AI solve new problems while it would be obviously much more efficient in helping to navigate the present body of works

At this point I use the models on many days; they were net negative back in August and are solidly net positive now.

Cheating here means claiming 2 when really something like 1 happened. It's not that they're not both meaningful, but they show pretty different behaviors and that's important. Oneshotting these problems totally autonomously (use many agents sure) is meaningfully different.

Re: #1stproof I just want to point out I think that 1) a math human with moderately relevant background can solve many of these problems by interactively talking to a model and then giving it hints, 2) this is extremely different from 1-shot performance, 3) it's tempting to cheat