Andres Vahter 🦀

@andres.vahter.me

My async compiler work got funded! So here's my first PR: github.com/rust-lang/ru... Roughly, it changes match x { true => foo(1).await, false => foo(2).await, } to match x { true => foo(1), false => foo(2), }.await This means the future will only have 1 state instead of 2

Add merge yields by diondokter · Pull Request #162429 · rust-lang/rust

Part of Async statemachine optimisation project goal r? dingxiangfei2009 Added a pass that merges functionally identical yields right before the async StateTransform pass. The StateTransform pass c...

github.com

How we developed the world's first safety-certified product written in Rust – and why we went bare metal > Sonair certified ADAR One, the first safety-critical embedded system in Rust, using Ferrocene on bare-metal TI Sitara for robot safety sensors.

Image: How we developed the world's first safety-certified product written in Rust – and why we went bare metal 


> Sonair certified ADAR One, the first safety-critical embedded system in Rust, using Ferrocene on bare-metal TI Sitara for robot safety sensors.

ok so it turns out discord is.. 80% there when it comes to interacting with agents. so many cool integration points. each thread even has its own identity, a-la pluralkit :)

My impression is that AI coding is on a "pick two of three" triangle: scope, ship speed, and correctness. Small tools are (speed + correctness) fast prototypes are (speed + scope) and big projects are (scope + correctness). Every project that tries to be all three degrades to "fast prototype."

Why is message passing better? Because it is closer to a purely functional style of programming, where messages come in and messages come out that are hopefully only determined by the state so far. You can build models for what should happen and then test the hardest parts

rain 🌦️@sunshowers.io · 3mo ago

The ability to do shared-memory concurrency imposes a massive burden on the ecosystem, because everyone keeps having to ask "is this dependency thread-safe". Most languages should just stick to message passing and the actor model (99% of the time, it's a better way to build software anyway)

How did we end up in a place where I now need to clean up terrible LLM code? I hate this. The original guy also hasn't learned anything. I'm 70% sure the PR review I gave was just fed back into the LLM, which means that *every single problem* had to spelled out in the review. Is this the future?

maybe im egoizing but it feels LLMs expose the “island of agency” where on one side you have people who are in some kind of slop fever (including most of the boosters on here) and on the other side you have people falling into superstitious denialism. in the middle you can get so much done.