Blog posts and notes from Leonardo de Moura — creator of Lean and Z3. https://leodemoura.github.io Mon, 20 Apr 2026 00:00:00 +0000 https://leodemoura.github.io/blog/2026-4-20-signal-shot-the-platform-is-ready/ https://leodemoura.github.io/blog/2026-4-20-signal-shot-the-platform-is-ready/ Mon, 20 Apr 2026 00:00:00 +0000 Blog Signal Shot is a public moonshot to verify the Signal protocol and its Rust implementation using Lean. Not the math on paper. The actual code. It is a joint effort of Signal (Rolfe Schmidt), the Beneficial AI Foundation (Max Tegmark), and the Lean FRO. BAIF's launch post is here. https://leodemoura.github.io/blog/2026-4-2-why-lean/ https://leodemoura.github.io/blog/2026-4-2-why-lean/ Thu, 02 Apr 2026 00:00:00 +0000 Blog Markus de Medeiros wanted to understand garbled circuits. He had no formal training in cryptography. He spent an afternoon in Lean: a circuit evaluator in functional style, a custom DSL via metaprogramming, OpenSSL bindings via FFI, a correctness proof for his serializer, and a mutable garbling algorithm in imperative style. All in one tool. His conclusion: "I genuinely do not know of another tool that can do all of the following." And then: "I fucking love this shit." That reaction keeps happen… https://leodemoura.github.io/notes/2026-3-28-cray-colloquium-at-the-university-of-minnesota/ https://leodemoura.github.io/notes/2026-3-28-cray-colloquium-at-the-university-of-minnesota/ Sat, 28 Mar 2026 00:00:00 +0000 Notes Next Monday, I am giving the Cray Distinguished Colloquium at the University of Minnesota Department of Computer Science & Engineering. The Cray Colloquium Series was founded more than 30 years ago with a donation from Cray Inc. to honor the contributions and legacies of Dr. Seymour Cray. Past speakers include Turing Award winners and members of the National Academy of Engineering. https://leodemoura.github.io/blog/2026-3-16-who-watches-the-provers/ https://leodemoura.github.io/blog/2026-3-16-who-watches-the-provers/ Mon, 16 Mar 2026 00:00:00 +0000 Blog In 2026, a skilled researcher using Claude (Opus 4.6), with real-time support from the Rocq development team, found seven kernel bugs in Rocq, a widely used proof assistant. Seven independent paths to proving something false. Rocq is a serious system with decades of foundational work. But even Rocq is vulnerable when AI applies sustained adversarial pressure to a single implementation. The question is not whether AI will find the bugs, but whether the architecture is built to survive them. https://leodemoura.github.io/notes/2026-3-15-this-website-is-now-powered-by-verso/ https://leodemoura.github.io/notes/2026-3-15-this-website-is-now-powered-by-verso/ Sun, 15 Mar 2026 00:00:00 +0000 Notes This website is now built with Verso, the documentation tool for Lean. Verso is developed by David Thrane Christiansen at the Lean FRO. It powers the Lean language reference manual, the Lean website, and now this site. The main motivation: Lean code examples in blog posts are type-checked at build time and include interactive hover information. Try hovering over the code in Teaching AI to Make Proof Automation Work. https://leodemoura.github.io/blog/2026-3-14-teaching-ai-to-make-proof-automation-work/ https://leodemoura.github.io/blog/2026-3-14-teaching-ai-to-make-proof-automation-work/ Sat, 14 Mar 2026 00:00:00 +0000 Blog Consider this goal: example {α} [CommRing α] [IsCharP α 0] [NoNatZeroDivisors α] (d t d_inv : α) (Δ40 : d * (d + t + d * t) = 0) (Δ41 : d^2 * (d + d * t - 2 * d * t^2 + d * t^4 + d^2 * t^4) = 0) (_ : d * d_inv = 1) : t + 2 * t^2 - t^3 - 2 * t^4 + t^5 = 0 := by grind https://leodemoura.github.io/notes/2026-3-13-sym---interactive-mode-for-scalable-verification/ https://leodemoura.github.io/notes/2026-3-13-sym---interactive-mode-for-scalable-verification/ Fri, 13 Mar 2026 00:00:00 +0000 Notes We are building sym =>, a new interactive tactic mode for writing scalable verification proofs in Lean. It extends grind's interactive mode with tactics from the Sym framework — our infrastructure for symbolic manipulation that achieves linear scaling on verification tasks where traditional approaches break down. https://leodemoura.github.io/blog/2026-2-28-when-ai-writes-the-worlds-software-who-verifies-it/ https://leodemoura.github.io/blog/2026-2-28-when-ai-writes-the-worlds-software-who-verifies-it/ Sat, 28 Feb 2026 00:00:00 +0000 Blog Code Metal recently raised $125 million to rewrite defense industry code using AI. Google and Microsoft both report that 25–30% of their new code is AI-generated. AWS used AI to modernize 40 million lines of COBOL for Toyota. Microsoft's CTO predicts that 95% of all code will be AI-generated by 2030. The rewriting of the world's software is not coming. It is underway. Anthropic recently built a 100,000-line C compiler using parallel AI agents in two weeks, for under $20,000. It boots Linux and c… https://leodemoura.github.io/blog/2026-2-18-proof-assistants-in-the-age-of-ai/ https://leodemoura.github.io/blog/2026-2-18-proof-assistants-in-the-age-of-ai/ Wed, 18 Feb 2026 00:00:00 +0000 Blog AI is transforming how we write formal proofs. Language models can now generate thousands of lines of verified mathematics with minimal human guidance, and the cost keeps dropping. This is exciting, and it raises a natural question: what role does the choice of proof assistant play when AI is doing more of the work? I believe the answer is that it matters more, not less.