Sebastien Bubeck ✓

@SebastienBubeck Look in the training data and confirm/deny that Buckmaster's data is or isn't in there. Until you do that, it's very very hard to trust anything OAI says or does. 74.1K views · 1.4K likes · 15 reposts · 10 replies Open on X →

Some more technical points: (a) We began working on the Millennium problems due to viral twitter rumors that Anthropic had resolved 2 Millenium problems. Our aim was to see whether our system was also capable of this impressive feat, especially given our excitement regarding 1M views · 2.2K likes · 103 reposts · 130 replies Open on X →

I would like to clarify a few things: 1) The screenshot is my reaching out to Levent to coordinate our releases. I hope it’s clear from the message that we came in with the best possible intentions. 2) I never ever asked for Levent to be removed from authorship of his own work

3.8M views · 5.7K likes · 486 reposts · 462 replies Open on X →

@SebastienBubeck Why didn’t you answer the question? https://t.co/TCkPV2HFGf

38.8K views · 764 likes · 10 reposts · 4 replies Open on X →

@SebastienBubeck Threats of ruining someone's career are famously known to be a part of the top-tier academic standards 26.9K views · 704 likes · 3 reposts · 3 replies Open on X →

@SebastienBubeck Which part are you disappointed with? The one where you threatened to end his career? 42.4K views · 982 likes · 4 reposts · 3 replies Open on X →

A series of false and inflammatory allegations against me are currently circulating on social channels. To clarify, I came into the discussion following academic norms, and I'm disappointed that it has come to this. Anyone who knows me knows that academic standards are of the 2M views · 2.3K likes · 138 reposts · 302 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck3 Sep

Meet Astra's TikZ unicorn. I find it mind blowing that a single entity can do this, as well as get essentially 100% on Frontier Math Tier 4, ARC-AGI 3, & ExploitBench. And it's not behind closed doors, anyone can just go and talk to it. Impossible to overstate the implications. 61.1K views · 557 likes · 37 reposts · 80 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck17 Aug

Astra?! 👀 292.5K views · 901 likes · 22 reposts · 20 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck13 Aug

This is amazing on so many levels, including the fact that I spent a week trying to solve this problem back in 2012 😅 46.9K views · 317 likes · 27 reposts · 10 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck12 Aug

Latest crazy story from the frontiers of math and AI -- a neurosurgery resident, with no training in advanced math, uses ChatGPT 5.6 to solve a major open problem in numerical linear algebra. https://t.co/xckiv4TCFc https://t.co/71ogXtzU9g 669.9K views · 3.1K likes · 497 reposts · 102 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck7 Aug

Anyone tangentially interested in AI should watch this talk. It gives a really good picture of what models today are capable of and what are the challenges we have to overcome. 54.8K views · 292 likes · 20 reposts · 8 replies Open on X →

@SebastienBubeck https://t.co/Eo9gUwgAYb 16.4K views · 280 likes · 5 reposts · 1 replies Open on X →

@SebastienBubeck https://t.co/NeCU9pPzBO 92.8K views · 2K likes · 124 reposts · 12 replies Open on X →

@SebastienBubeck Crazy how all of this was started by one guy bullying his Claude during the World Cup final. https://t.co/THCiql9ZZq 22.4K views · 213 likes · 4 reposts · 6 replies Open on X →

yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model. We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann 4.2M views · 6.7K likes · 947 reposts · 276 replies Open on X →

Sebastien Bubeck ✓@SebastienBubeck21 Aug 2025

@SebastienBubeck okay we're going to play a game - i'm going to guess what this guy does before looking at his profile. he works at an AI company and he tries to pitch AI. how close did I get? 15.8K views · 463 likes · 7 reposts · 9 replies Open on X →

And yeah the fact that it proves 1.5/L and not the 1.75/L also shows it didn't just search for the v2. Also the above proof is very different from the v2 proof, it's more of an evolution of the v1 proof. 136.8K views · 531 likes · 14 reposts · 22 replies Open on X →

By the way this is the proof it came up with: https://t.co/Ct8UBqKAHj 121.1K views · 481 likes · 29 reposts · 21 replies Open on X →

In the v1 of the paper they prove that if eta is smaller than 1/L (L is the smoothness) then one gets this property, and if eta is larger than 1.75/L then they construct a counterexample. So the open problem was: what happens in the range [1/L, 1.75/L]. 129.8K views · 258 likes · 5 reposts · 1 replies Open on X →

Now the only reason why I won't post this as an arxiv note, is that the humans actually beat gpt-5 to the punch :-). Namely the arxiv paper has a v2 https://t.co/kAO0XWNfyh with an additional author and they closed the gap completely, showing that 1.75/L is the tight bound. 117.4K views · 423 likes · 16 reposts · 7 replies Open on X →

As you can see in the top post, gpt-5-pro was able to improve the bound from this paper and showed that in fact eta can be taken to be as large as 1.5/L, so not quite fully closing the gap but making good progress. Def. a novel contribution that'd be worthy of a nice arxiv note. 118.6K views · 305 likes · 4 reposts · 1 replies Open on X →

The paper in question is this one https://t.co/q3r2LD0tPq which studies the following very natural question: in smooth convex optimization, under what conditions on the stepsize eta in gradient descent will the curve traced by the function value of the iterates be convex? 147.8K views · 447 likes · 15 reposts · 5 replies Open on X →