Are you interested in software reliability, software verification, programming languages, or formal methods? Would you like to work with me and my group in beautiful Lisbon? Get in touch to discuss available positions. See https://t.co/FCsrNd2gpo for more details.
entry-level jobs were never just about the work. they were where you learned how to do the work.
if AI takes that first rung, the learning doesn't disappear. it just stops happening by default. you have to go get it.
People thought I was crazy when I wrote “Computing without Processors” in 2011.
I’ve long dreamed of one way to program everything that computes. You’d write a single program and compile it automatically to any mix of hardware and software, with the split shifting to fit your performance and cost targets.
That’s been hard. Hardware and software use different languages, semantic models, and abstractions. The people who build them come from different university departments with very different cultures. These divisions exist because of how humans organize themselves, and because of the limits of human brains.
AI doesn’t care about any of this turf marking. It relentlessly synthesizes hardware and software together, integrates them, and formally verifies the result, or makes it correct by construction by deriving it from a formal spec. In some cases we may not need CPUs at all: an application’s control logic gets “partially evaluated” and dissolved into the computation itself.
I’m hoping to get the last laugh, probably sometime in 2027.
https://t.co/B0E8htUWzU
Veil 2.0 is out: verify all your distributed protocols in Lean 4.
New: verified model checkers for safety (2–20x faster than TLC) and liveness, LTL proof mode, TLA+ → Veil LLM skill, PlusCal-style DSL.
Great work by @zqy1018 and external contributors!
https://t.co/9OMsf1Kxzt
Recently, companies like @HarmonicMath, @andtruth, @SfTensor and @theoremlabs have all been betting on a full stack verification solution (going from the algorithm implementation down to the machine code). What's going on there? New blog post at
https://t.co/gi5l4y3vOp
Just to make sure that there isn't any misinformation among authors - the following post mentions "all ACM journals". This is not correct. TOSEM will continue to remain single anonymous for now.
I'm thrilled to bring &Truth out of stealth as its co-founder and Chief Science Officer. With a team I'm proud to be part of, we're setting out to specify and prove all the world's software correct and secure.
If you geek out over code people can understand and trust, join us!
Primo is now open source: an LLM-built SMT solver with published results ahead of the SMT-COMP 2026 QF_LRA.
Want to help? Try it, report bugs, improve the docs or contribute code. Questions welcome too!
https://t.co/6dmCu1dHXU
#SMT#FormalMethods
Did your #ICSE2027 paper not make it this time? Don’t let good research stop there. 💡
Revise and submit your Formal Methods + Software Engineering paper to #FormaliSE2027!
📍 Dublin, Ireland | 26–27 April 2027
📝 Deadline: 6 Nov 2026 (AoE)
#FormaliSE2027#ICSE2027#FM#SE
We should maximize the amount of proof we can get out of AI while the money cannon is on. It's like early days of Uber, except the benefits are 100x more tangible. A fully-verified hw+sw stack? That's like gold everyone can share w/out having to divide. Generational opportunity.
📦 Can AI agents automate software packaging and maintenance?
Excited to share Vibenix, our new open-source AI agent that automatically generates and maintains reproducible Nix packages!
Instead of relying on one-shot LLM generation, Vibenix iteratively builds software, analyzes failures, repairs packaging expressions, and validates the results.
🚀 Some exciting results:
✅ 95.3% build success across 466 real-world packaging tasks, compared to just 5.6% for a single-shot baseline.
✅ 47% → 82% functional correctness after an additional runtime refinement stage.
✅ 87.6% success in automatically maintaining existing packages.
We also introduce NixBench and NixBench-Repair, two new benchmarks with over 1,200 tasks!
💡 An important finding: just because software builds successfully doesn't mean it works correctly! Runtime validation makes a substantial difference.
Our results show how combining LLMs with execution feedback can make AI agents more effective at real-world software engineering tasks.
📄 Paper: https://t.co/6dkJbFW01S
#AIAgents #SoftwareEngineering #Nix
Lean v4.35.0 is out, and it ships with verified kernels.
Run `lake check --paranoid` to recheck your project with every bundled kernel.
Blog post: https://t.co/c35WKawLzN
Alfréd Rényi famously said: "A mathematician is a machine for turning coffee into theorems."
2026 update: "A mathematician is a machine for turning tokens into theorems."