2026-03-17 12:03 UTC
"Code should be written to be verified, and only incidentally to be executed"
Proposition: Code written in functional languages is both intrinsically and extrinsically more verifiable. Intrinsically because of idempotency. Extrinsically because social forces happen to have resulted in code written in languages like Elixir to contain patterns that are easier to model.
The attached opinion piece suggests this makes them better languages for LLMs. But that also makes them better languages, period.
https://www.thefuriousopposites.com/p/grace-hoppers-revenge
👋🏽 @cdegroot
Replies (1)
-
@cdegroot@mstdn.ca 2026-03-17 13:14
@raganwald we're using Claude Code at work. It's remarkably good at writing Elixir code. It is also remarkably bad at design, and if you ever manage to accidentally blow through its window size, oh boy you're in for a surprise. And needless to say, production/performance constraints usually don't enter the picture. It's more of a code typing assistant and simple refactoring assistant than a coding assistant at this point. And my prediction is that we've reached a point where progress is gonna be, well, progressively harder. tl&dr: a somewhat dim intern ;)