Ilya Sergey

New blog post on Proofs and Intuitions by @zqy1018: Liveness Proofs in Veil, Part I: The First Step. proofsandintuitions.net/2026/06/24/liv… Safety says nothing bad happens; liveness says something good eventually does. We present a proof mode for verifying liveness deductively in Lean.

Industry is where academics go to do one thing at a time.

What I find fascinating with Claude Fable 5 is it proves once again that large generalist models will outperform vertical ones. On ProofBench (graduate-level formal math benchmark in Lean, where a proof either compiles or it doesn't) Fable 5 beat Harmonic's Aristotle, 77% vs 71%. Aristotle is a system built specifically for formal math + run on its own internal harness, so the generalist beat the specialist on the specialist's home turf. It's the Richard Sutton's "The Bitter Lesson". His whole argument is that across 70 years of machine intelligence research, the methods that win are the general ones that scale with compute. Not the ones where we hand-encode human expertise. Building our own knowledge into the system feels good and helps short term gains but long term it always gets overtaken by bigger model. You can look at Chess, Go, speech, vision, same story every time. First the specialized model wins, then the general one takes over. and btw this is the whole premise of AGI. You don't build one model for math, one for code, one for law. you build a single general model that scales with compute and it learns to do everything

@headinthebox Looking forward to tell you why you are wrong (spoiler: it's about metaprogramming).

My statement for FMxAI 2026 next week.

Can you identify this city without Googling?

@satnam6502 @headinthebox @tthomasdd Hey, all my submissions have heen rejected from the Fine Dining track at PLDI’26, but will be at the conference anyways. Happy to set up a parallel track.

@headinthebox @tthomasdd It turns out there is no room for me at the dinner I want to crash so perhaps I will not be flying in after all.

Say you are interested in programming languages and want to attend pldi26.sigplan.org to hear about all the exciting research results from your academic friends: * Registration fees: $1,340+2*$450 * Accommodation: 7*$250 * Food: 7*$75 * Travel: $500 (within USA) That is over $5000 just for the privilege; or you can buy 6 base model Mac minis, or pay the rent.

@samth @satnam6502 @headinthebox Don’t waste your time arguing.

@satnam6502 @headinthebox Conferences are expensive because catering is expensive.

Interesting discussion on LinkedIn regarding the no-AI-review policy we instituted for OOPSLA'27. The opponents' main argument: "peer reviews by humans mostly suck anyway, so by actively using LLMs to write reviews we won't lose much in quality, but will save everyone time".

@ShriramKMurthi This was a great read. Thanks for sharing!

This is a beautifully-written piece and extremely accessible to anyone just with mathematical maturity. Highly recommended. Crystallizes so many things that have been jostling around in my head in terms of work I've been looking for (around the Kolmogorov intuition).

New blog post: On the Unreasonable Effectiveness of Property-Based Testing for Validating Formal Specifications. proofsandintuitions.net/2026/05/18/pro… The gist: randomised testing can validate formal specs. It's very cheap and powerful: we found bugs in specs of VERINA and CLEVER benchmarks.

Many people have claimed that with AI-assisted bug finding, secure code (and hence trustless anything) will be impossible. I have a much more optimistic take, and AI-assisted formal verification is a major part of the reason why: vitalik.eth.limo/general/2026/0…

Causal consistency is hard.

@HerrDreyer Giving a hard time about basic typing rules?

Come to MPI-SWS, and you too can stand around with me at a whiteboard I never use while I lecture you about basic typing rules! That's how research works, you know? (My students won't stop giving me a hard time about this... 🤣) linkedin.com/posts/were-lau…

I've sent my paper draft to colleagues for feedback. Every comment I got was amazingly informative and constructive. Each one was also absolutely idiotic. All of them were pretzels. And somehow, every last piece of feedback I got was a small dog named Mortimer.

@lastland0 @xuanalogue My meetings are side effect-free.

celebrating post-deadline just now! 3 NeurIPS submissions, first big conference push since starting as faculty -- we'll see how things turn out but I'm very proud of the students 😊

@AndrewCMyers By building tools, of course!

@ilyasergey How do we get the gold out there where it is useful to others?

Programming language research operates a gold mine. Adding gold is encouraged, even (and especially) when unrefined. Borrowing is permitted if you return more than you took. Refining it and spending on something useful counts for little. This is why the gold stays in the mine.

11. Last but not least, George Pîrlea's @GeorgePirlea talk on Veil: Multi-Modal Verification of Transition Systems ...and this is done with Lean @leanprover ! youtube.com/watch?v=24mMfU…

It was great to attend the #tlaplus community event this year and showcase our work on Veil and its new concrete state model checker, Lace. Thanks to the organisers for the invite!