Haskell exits production after 7 years, vibe-coding blamed
Scarf posted a detailed writeup explaining why they moved away from Haskell after seven years in production. The most memorable comment on HN was blunt: the post is a lot of words for 'we are vibe-coding our app now and the Haskell compiler is too slow for the agent.' That is a sharp read.
The pattern across this thread and others today is the emergence of a real question about language choice in an AI-assisted coding world. If your primary code author is an LLM agent, the calculus for choosing a language shifts. Haskell's expressiveness and type-system guarantees matter less if the agent is the one reasoning about correctness, and they matter more if the agent makes mistakes and you need the compiler to catch them. People are genuinely unsure which direction to run.
One commenter framed it as being in a 'post-language world.' That is probably too strong, but it points at something real: the era of choosing a language primarily for its developer experience is being disrupted by agents that write most of the code and care nothing about ergonomics.
So what?
If you are choosing a stack right now, factor in how well your language toolchain works with AI coding agents. Haskell's compiler is a great bug-catcher for humans but a bottleneck for agents. Python and TypeScript win on ecosystem support for LLM tooling even if they lose on correctness guarantees.