My main gripe is that the syntax is too sugary. I've often wondered to myself what a Haskell with syntax more in the spirit of Python would look like. I just wish I didn't always feel like I had to literally memorize what various custom operators meant with flashcards or something. Also, Haskell has relatively complex parsing and evaluation rules. The meaning of well-written code shouldn't be obscured by syntax.
And, along the lines of the syntax, the culture of style in Haskell. Single letter variables abound. Not just throwaway vars where the meaning is pretty clear from context. Entire libraries written with only single letter or acronymic variable names. Good luck understanding what a library author was trying to do by reading their code. I think this culture problem is largely being driven by the fact (which I mentioned) that Haskell's base syntax is already very opaque. I understand that the language takes some cues from the mathematical community (not surprising given its history). However, the language needs to ditch this to gain broader acceptance.
And, again, leading into my next point: the culture of the community in general is too academic. I'll make another comparison with Python. The Python community, among scripting languages, leans more toward the academic side but in a good way. They've found a nice balance. It's easy to get started and, if you want to learn more about the details, you can do so. With Haskell, you immediately get hung up on the details. Most of the details have to do with how Haskell fits into the framework of category theory. However, most programmers don't give a shit about provability. They just want to write software that is faster, safer, and more concise without being terse. I'm not convinced that it's impossible to have this in a functional language with the same goals as Haskell. Beginners should be able to get started quickly and without much confusion. If they're curious about the underlying principles, they should be able to gradually peel back the layers and learn more.