GoKawiil
Tech News
Lasm is a new toolchain that ahead-of-time compiles standard Lean 4 programs into WebAssembly, letting a single codebase run as a Node server, serverless or edge function, WASI/Fastly module, or browser app. It ships via npm bundling Lean and its Lake build tools, requiring only Node, Bun, or Deno. Its flagship demo, a RealWorld-spec app called Conduit, includes 117 formally proved theorems.
Ben Lorica interviewed Leo de Moura, creator of the Lean theorem prover and VP/Distinguished Scientist at AWS, about how AI is making formal verification more accessible. The conversation covered machine-checked proofs, trust in AI-generated code and autonomous agents, and the concept of a fully verified software stack.
OpenAI and Anthropic both announced AI-generated proofs this week that refute the exponential time hypothesis, a foundational assumption in theoretical computer science used to explain why certain problems require exponentially more effort as they grow larger. The proofs, published on GitHub in natural language and machine-checkable Lean format, have not yet been independently verified by researchers. OpenAI has reached out to the hypothesis's original proponents and other academics to help review and validate the findings, including funding a dedicated conference.
A research team has produced a complete machine-checked formalization in the Lean proof assistant of the classification of finite simple groups, following the original first-generation proof. The formalization comprises over 43 million lines of Lean code and 2.05 million declarations, drawing on 715 papers and books. The team also reports finding an issue in a previously published argument for the Unbalanced Group Theorem and describes the alternative proof route they used instead.
A developer spent time optimizing roc-deflate, a DEFLATE compression package for the Roc functional programming language written by Niclas Ahden, aiming to see how close it could get to the performance of libdeflate, the C library widely regarded as the fastest DEFLATE implementation. The effort followed an earlier result showing a Lean 4 library, lean-zip, beating Rust's popular miniz_oxide package but still falling well short of libdeflate, which uses hand-tuned, architecture-specific SIMD code. Testing used the non-streaming APIs and the Silesia corpus, since roc-deflate does not yet support streaming.
Sheryl Sandberg's organization Lean In released new survey data showing that six in 10 employees say they've encountered sexist opinions at work that they trace to manosphere and tradwife media influences. The findings suggest online gender-role content is bleeding into workplace culture and conversation.
A group of mathematicians and formalization specialists, including Jeremy Avigad and Kevin Buzzard, published a guest post responding to calls—spurred by a recent OpenAI release—to use AI to formalize all of mathematics quickly. They say they are optimistic about AI's potential but worried that a rushed, uncoordinated effort could waste resources or cause lasting harm, and they outline conditions for doing large-scale formalization well.
The Lean proof assistant's new v4.35.0 release introduces a lake check command that re-verifies developments using leanchecker inside a sandbox, plus a --paranoid mode that runs four additional kernels: nanoda, lean4lean, con-leche, and con-ron. Con-leche, built in Lean itself by Claude models under supervision from Joachim Breitner at the Lean FRO, is formally proven consistent against a set-theoretic model and reportedly runs faster than the official kernel on the full Mathlib library.
The Simons Foundation announced a partnership with Anthropic and OpenAI committing $15 million in combined funding and resources to help mathematicians adapt to AI's growing role in research. The foundation will manage the money, including an open call for proposals, funding workshops and initiatives to evaluate AI-assisted results, train researchers on new tools, and develop shared standards. Anthropic and OpenAI will also supply AI tools and computing resources to foundation-supported researchers in mathematics and basic science.
A developer created a complete implementation of RealWorld, the Medium-clone reference application used to benchmark full-stack frameworks, using only Lean 4, a functional language best known as a proof assistant. The project includes a server, database integration, and frontend rendering, and passes RealWorld's official API and website test suites. The frontend uses a library called datastar-lean for interactivity, and the full source code is published on GitHub.
On October 6, OpenAI released more than 700 mathematical papers produced by an internal AI system, some with formal proofs written in the Lean proof-verification language and others still at earlier stages of verification. A mathematician who reviewed the release, previously a FOCS program committee chair, reported feeling detached from the material rather than engaged, unlike his experience evaluating human-submitted research.
OpenAI's September 8 announcement of a solution to the Navier-Stokes problem included two versions of its proof: one in natural language with mathematical symbols, and one formalised in the Lean programming language for automated verification. A team led by Anders Hansen and Fabian Circelli at the University of Cambridge says these two versions do not actually match, meaning the Lean code may not faithfully represent the natural-language argument.