
Computer Scientist / CS Prof at University of Porto / Researcher at INESC-ID (@InescID) / Coding with https://xitter.cf/t.co/IRUJxE7E9D (@codesetai)
Lisbon (PT)
Joined April 2007
- Tweets4.2K
- Following719
- Followers993
- Likes9.4K
João F. Ferreira retweeted
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
João F. Ferreira retweeted
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.
João F. Ferreira retweeted
Hoping a formal methods expert becomes an influencer before an influencer becomes a formal methods expert.
João F. Ferreira retweeted
It's great to see more and more folks starting to use TLA+. @lemmster
we've been trying this approach of skill based generating TLA+ at @datadoghq for quite some time.
wrote a few blogs on it. some of my team affectionately made a sesh-mode skill in one of our oss repos. Added all links in comment.
We now have something more advanced, will share soon.
a meta-skill that generates the whole suite is a great idea. antithesis skills do something similar for DST, research properties > workload > triage
João F. Ferreira retweeted
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: leodemoura.github.io/blog/20…
João F. Ferreira retweeted
New paper about how we converted English documentation into Liquid Types using a DFA-guided agentic workflow: wiki.alcidesfonseca.com/rese…
João F. Ferreira retweeted
Mystery Function 🔥 1 7/7
October 7, 2026
f(⚫ ⚫ ⚫ ⚫ ⚫ 🔴 🟢)
codeset.ai/function/bmMqb
João F. Ferreira retweeted
Three months, two AI agents, and 5 million lines of Lean code later, delighted our classification of all order-6 semigroups is heading to MATH-AI@NeurIPS 2026!
Major announcement in mathematical formalization!
Finally, after 3 months of very intensive, nonstop work by several AI agents (Codex and Claude Code), we have settled a classification of ALL 15,973 semigroups of order 6. Every one of the 15,969 that has a finite basis now has it written down explicitly, and the remaining 4 are proved to have none at all.
The whole list of bases is produced for the first time in history. Until now, for most of these semigroups, mathematicians only knew that a basis exists; nobody had ever written one down. The formalization in Lean, orchestrated with a multi-agent approach, reached more than 5 MILLION lines of verified Lean code, one of the largest auto-formalization projects to date.
It's been an amazing joint work with @JanotaMikolas and @Jan_Hula, accompanied by senior experts in semigroup theory, João Araújo and Edmond W. H. Lee.
Our paper describing this project, "Proving at Scale for Universal Algebra", has been accepted at the MATH-AI workshop, NeurIPS 2026!
When we launched this project, we were a bit pessimistic: producing proofs for 15,973 semigroups seemed out of scope for the current technology. But with a careful setup (agents communicating through a mailbox we arranged, orchestrating the effort with a custom method of bootstrapping and auto-research), it has finally come to the finish line. The last 390 semigroups took 23 more days. The very last one needed a whole structural analysis before its proof could even be written.
HotCRP had experienced a system-wide failure a few minutes before the FSE'27 submission deadline.
To give all authors a fair chance to finish their submissions, we've re-opened the submission site until October 3, AoE, i.e., 24 hours later than the original deadline.
🤖 Made with AI
João F. Ferreira retweeted
I'll present TLAPS-Bench at the Frontier Data Summit (thanks @SnorkelAI @_Incynthia), in the hope to let people know about the benchmark and find resources to run it.
TLAPS-Bench is a challenging and long-horizon proof benchmark. The TLA+ specs in the benchmark come from real-world systems and protocols, so proving them has great practical values. It's expensive and still hard for frontier AI to do autonomously. We hope that with TLAPS-Bench and its toolchains, every important TLA+ spec can be formally proved beyond MC.
The benchmark is a team effort of the @tlaplus foundation, the #Specula team, and many collaborators. While I'm physically the closest to SF, @lemmster initialized the project and @qiancheng9788 did the major technical work; Ruize Tang, @SaadMRP @lemmster Stephan Mertz, Igor Konnov, Chris Newcombe, @Yiming_Su3 @MunimThahmid MR Sukanya, @xu_dong_sun @disalg_spec @HacksonClark all contributed in different ways.
Special thanks to @leochen_ai and @JiaweiLiu_ for supporting the benchmark, without whom we would not survive and see the unexpected interests in TLA+.
For AI companies, we'd love to run your models/agents if you could throw us some tokens.
TLAPS-Bench uses the TLA+ Proof System to formally prove the correctness of complex protocols and systems. It also provides a rich, extensible, and scalable benchmark suite for frontier AI. From @tianyin_xu, one of 25+ posters at Frontier Data Summit.
frontierdatasummit.ai/
João F. Ferreira retweeted
Excellent work on formally verifying the #Kubernetes control plane using a compositional approach led by @Cat_overflow and @xu_dong_sun, with the awesome team of @NikhilDate5 @tylergu_jiawei Cody @tchajed and Oded. It shows a practical path to verify a large control plane by gradually verifying individual controllers progressively.
The compositional verification is enabled by a new spec termed CORE (COmpositional REconciliation), which reuses ESR (the original #anvil spec) to specify each controller's reconciliation and employs rely-guarantee conditions to restrict inter-controller interference. CORE encodes both liveness and safety -- liveness captures correct state reconciliation logic and safety restricts inter-controller interference.
Check out the paper which has a great amount of insights on specification, proof, and implementation.
cathy-cai.page/pubs/welder26…
João F. Ferreira retweeted
Replying to @kristianfreeman @spacexai
Checkout the codeset team: linkedin.com/posts/codesetai…
João F. Ferreira retweeted
Vibecoding brain test #1
You have 10 seconds.
What does this print?
A) [1] then [2]
B) [1] then [1, 2]
C) [1, 2] then [1, 2]
D) TypeError
João F. Ferreira retweeted
We made a meme quiz to test your post-AI SWE skills
Can you still code without an AI?
codeset.ai/fried/swe
João F. Ferreira retweeted
Verus is an open-source, automated program verifier for Rust that mechanically checks code against a formal mathematical specification for all possible inputs. Amazon used it to prove correctness of Nitro Isolation Engine primitives. amazon.science/blog/developi…
João F. Ferreira retweeted
It's not an unfair statement. On a separate note, it's a time we really should help our junior colleagues who are starting their faculty jobs. Senior folks (including me) should do more on mentoring and collaboration. University administrations should understand the challenges and support international faculty and students, instead of falling into the government propaganda.
João F. Ferreira retweeted
I'm proud to officially announce CUE Labs!
I co-founded CUE Labs with @_myitcv to build a Configuration Control Plane, solving config chaos with CUE. Our work also provides the stewardship for the @cue_lang open source project.
Read our launch post: cue.dev/blog/announcing-cue-…
João F. Ferreira retweeted
Now for our latest books give-away: 38 Classics books to be raffled off for free. Just RT this message, follow us, and we will make the draw on Monday 9 June. 1st name gets 20 books, 2nd 12, 3rd 6. Open to all, good luck!
(Will post details and photos of each book this evening.)























