A bit off-topic, but I see places like Tweag claim to use Haskell in production for lots of things, including domains where I would never consider this language a good fit, like statistics & machine learning.
I'm very familiar with the Lisp family of languages (Scheme, Common Lisp, Clojure, Dylan, Julia...) and with some from the ML family as well (SML and Scala).
Is Haskell worth the effort for a small organization like a startup or a research lab? Would it bring something different when compared to those other languages? If so, in what domains? If one wants to have strict guarantees in the form of proofs why not using Agda or Idris instead?
My experience with Haskell is less “strict guarantees with proofs” and more “changes to state are exposed in the type system.” I also think that Haskell has a much larger ecosystem than Agda or Idris, so if you need libraries you can find them.
>If one wants to have strict guarantees in the form of proofs why not using Agda or Idris instead?
Because they're research languages. The amount of non-toy software actually verified in them is zero. I'm not even sure if there's any non-toy software written in Idris besides the Idris compiler itself.
Coq's track record is better, but its language is ugly as hell in comparison (and to actually do anything you'd need to extract to Ocaml/Haskell anyway and the same applies to Agda).
I'm very familiar with the Lisp family of languages (Scheme, Common Lisp, Clojure, Dylan, Julia...) and with some from the ML family as well (SML and Scala).
Is Haskell worth the effort for a small organization like a startup or a research lab? Would it bring something different when compared to those other languages? If so, in what domains? If one wants to have strict guarantees in the form of proofs why not using Agda or Idris instead?