July 10, 2026
Compile times? More like crime times
Scarf has moved away from Haskell
AI fever hits Haskell as fans split over Scarf’s dramatic breakup
TLDR: Scarf says it’s leaving Haskell because AI-assisted coding changes the math: slow build times now hurt more than ever. Commenters are split, with some calling that sensible and others saying AI’s messy code makes strong safety checks even more valuable.
Scarf, a company that built big parts of its business on Haskell, has now publicly said it’s moving away from the language — and the internet immediately turned it into a messy family argument. The company’s founder made it clear this wasn’t a casual fling gone wrong: he’s been a Haskell believer for 16 years, helped lead parts of the community, and says the language delivered real reliability in production. But in the age of AI code helpers, he argues, long build times and setup pain now feel like a dealbreaker. In plain English: if an AI can spit out code in minutes, waiting ages for the computer to check it starts to feel unbearable.
And wow, the comments did not quietly nod along. One camp basically yelled, “Are you kidding? AI makes strong guardrails more important, not less,” with several people saying they trust stricter languages because AI often produces junk. Another camp mocked the whole thing as a glorified “vibecoding” confession — a spicy way of saying Scarf wants to let AI guess its way through coding faster, even if the code is sloppier. Others zoomed out and wondered if we’re entering a “post-language world,” where the best tool is simply the one that’s fastest for both humans and bots. Meanwhile, some readers were baffled by the move toward Python, arguing it’s exactly the kind of language AI makes look worse, not better. In short: one company changed tools, and the community reacted like someone announced the breakup at Thanksgiving.
Key Points
- •Scarf says it has moved away from Haskell after using it as the foundation of its backend since launch.
- •The article states that Scarf successfully ran production Haskell systems with uptime requirements and contractual SLAs for years.
- •Scarf’s Haskell stack included a main API built with Servant, Beam, and PostgreSQL, plus a high-performance Scarf Gateway service built on WAI.
- •The author says Haskell provided reliability, bug detection through its type system, strong domain modelling, and good performance.
- •The article argues that AI coding tools changed the cost-benefit balance by making compile time, cold builds, cache setup, and parallel agent workflows more important bottlenecks.