We are organizing FAIRS: Foundations of AI Rising Stars (warwick.ac.uk/fac/sci/dcs/ev…). If you are a PhD student or postdoc working on foundational aspects of AI and is/will soon be on the job market, please apply. Deadline is 15th July.

Thanks to @sigact for supporting this event!

Rare to be in an era where there's an elephant in the room looming over nearly *every* conversation!

Davin Choo is joining the NUS faculty @NUSComputing : linkedin.com/feed/update/urn…. Go work with him! Davin is beyond fantastic!

In my doctorate, I proved the Erdős Primitive Set Conjecture, showing that the primes themselves are maximal among all primitive sets. This problem will always be in my heart: I worked on it for 4 years (even when my mentors recommended against it!) and loved every minute of it. [Primitive sets are a vast generalization of the prime numbers: A set S is called primitive if no number in S divides another.] Now Erdős#1196 is an asymptotic version of Erdős' conjecture, for primitive sets of "large" numbers. It was posed in 1966 by the Hungarian legends Paul Erdős, András Sárközy, and Endre Szemerédi. I'd been working on it for many years, and consulted/badgered many experts about it, including my mentors Carl Pomerance and James Maynard. The the proof produced by GPT5.4 Pro was quite surprising, since it rejected the "gambit" that was implicit in all works on the subject since Erdős' original 1935 paper. The idea to pass from analysis to probability was so natural & tempting from a human-conceptual point of view, that it obscured a technical possibility to retain (efficient, yet counter-intuitve) analytic terminology throughout, by use of the von Mangoldt function \Lambda(n). The closest analogy I would give would be that the main openings in chess were well-studied, but AI discovers a new opening line that had been overlooked based on human aesthetics and convention. In fact, the von Mangoldt function itself is celebrated for it's connection to primes and the Riemann zeta function--but its piecewise definition appears to be odd and unmotivated to students seeing it for the first time. By the same token, in Erdős#1196, the von Mangoldt weights seem odd and unmotivated but turn out to cleverly encode a fundamental identity \sum_{q|n}\Lambda(q) = \log n, which is equivalent to unique factorization of n into primes. This is the exact trick that breaks the analytic issues arising in the "usual opening". Moreover, Terry Tao has long suspected that the applications of probability to number theory are unnecessarily complicated and this "trick" might actually clarify the general theory, which would have a broader impact than solving a single conjecture.

Human insight is still a thing: over the last few years many computing resources were thrown towards the Merzon-Smirnov conjecture on maximal Schubert polynomials, including DeepMind's FunSearch. In the end it fell to a human-generated targeted check: arxiv.org/abs/2603.20104

Eerie...

.@SebastienBubeck is so so good; incredible researcher and leader.

Me defending my O(n^3) solution to the coding interviewer.

I know nothing about the NS claims by @davidmbudden, but this comment strikes me as true. The fact that a Lean program compiles sans errors says nothing about whether a theorem has been proved...unless a human verifies that the program encodes a valid proof.

We need more senior researchers camping out at their posters like this. Managed to catch 10 minutes of Alyosha turning @anand_bhattad’s poster into a pop-up mini lecture. Extra spark after he spotted @jathushan. Other folks in the audience: @HaoLi81 @konpatp @GurushaJuneja.

I finally understand RL! 😁

It's beginning.

Seeking Ph.D. candidates with research interests in theory and foundations of AI & ML: warwick.ac.uk/fac/sci/dcs/re…. 🏁Apply by 14 November 2025

Proud advisor moment! My Ph.D. advisee, Davin Choo (cxjdavin.github.io/), postdoc at Harvard for a year now, got 4/4 NeurIPS accepts (including a spotlight). Look out for him at San Diego!

Interesting work by Davin (cxjdavin.github.io/) and collaborators, applying the lens of learning-augmented algorithms to improving healthcare coverage in Ethiopia: arxiv.org/abs/2509.00135.

🚨 Submissions for ALT2026 are now open! Submit your best work and join us in Toronto!🚨 Deadline: October 2 AoE Conference: Feb 23-26, 2026 algorithmiclearningtheory.or…

Dear Reviewers, It is completely fine to admit you were wrong in your initial evaluations. You will not lose anything, and the authors, your AC, and your SAC will appreciate your intellectual honesty. Best, One of the SACs