The reason Lean succeed so much and other theorem provers failed is, as much as the types crowd hate to hear it, because they invested in metaprogramming. Type theory ppl care about the beauty of metatheory, they don't care about user experience, and that's why most other provers are a pain to use.

Show Post

microsoft is more willing to give you money if you are willing to court users; leo if nothing else deserves kudos for doing the legwork a lot of theorem prover ecosystems often feel like they don't care at all about outreach, build it and the users will come, which simply isn't true

Show Post

This is also we're seeing a renewed interest in formal methods—ie, mathematical proof of software correctness. These have long been regarded as too much of a pain in the ass, but if LLMs can help you do the grunt work of writing the code proving the invariants you define, that changes the equation.

Show Post

[new blog post] Composition and Modular Verification of TLA+ specs muratbuffalo.blogspot.com/2026/08/comp...

Show Post

a big part of why I like the formal verification approach I've been taking is that I don't have to read the code, I just have to read the proofs there are still a lot of proofs, but they're only about a fifth of the entire codebase by loc

Show Post