Brett Evans

Brett Evans

@brettevans.me

Functional programming, dependent types, denotational semantics, category theory, cooking, snowboarding, rugby.

376 followers2.1K following65 posts

LLM prompts with “make no mistakes” is the funniest copium and should disqualify anyone using it from prognosticating on the ability of current AI.

With Mythos decreasing the cost of exploits, formal verification will move from “nice-to-have” but too expensive to the new minimum standard.

There needs to be an alternative teaching of computer science. One where the semantics of a language and construction thereof is taught before its syntax and parsing. One where algorithms is algebra and only later do you learn about representation, memory, and performance.

I have had the same thoughts while working with Agda. Use LLMs for proof search. I've been revisiting Bird and Meertens' work with this new lens.

Martin Kleppmann@martin.kleppmann.com

New blog post! ✨ I argue that AI will make formal verification go mainstream. martin.kleppmann.com/2025/12/08/a... Three reasons for this: 1. LLMs are getting increasingly good at writing proofs using proof assistants. This will make formal verification vastly cheaper than it's been to date.

Brett Evans reposted

Worth a watch: Head of Signal, Meredith Whittaker, on so-called "agentic AI" and the difference between how it's described in the marketing and what access and control it would actually require to work as advertised.

Great knowing VCs purchased their way into the GOP to gut the CFPB and SEC, and create a federal crypto reserve so the American public can bail them out of their crypto investments.

Fuzzy feeling knowing that billionaires believe the wealth they’ve accumulated is evidence of their great capital allocation skill and is deserved.

Nathie@nathievr.bsky.social

This is Meta’s brand new ad for Horizon Worlds. Not sure how I’m supposed to feel about it.

So now that tech bros have captured the U.S. government we are now subject to learning why federal agencies exist from “first principles” instead of, you know, reading about it.

In a society where access to higher education isn’t guaranteed the lie-to-children teaching method really rears its ugly head in political discourse. It’s frustrating to have to confront beliefs built upon elementary knowledge without also sowing distrust in experts.

Unsubscribe from email flows are truly a product backwater. With all the edge cases and lack of motive to support you’re lucky if you ever get unsubscribed succesfully.

You are up to date.