For me too there is a feeling of poetry... but for me it's more the nails-on-blackboard grating of awful poetry, and the ice-pick-to-ears screeching of abrading subway wheels. 🙂
Lisps merely inspire me towards elegant craftsmanship. But Haskell tempts me to dream of more. It's my fault, but ouch.
Dreams of math made manifest. Of materializing a lattice of theories, each with types and operators and laws, simply extensible and composable - a push-out lattice. Dreams of programming that's more math than plumbing. And maybe someday, some future Haskell Next will permit that flight, that dance. But that's not Haskell.
Someone I know, will pull up the deep math of a problem domain, model it in Coq, and mentally compile it to Haskell, extending GHC internals as needed to better serve as compiler target. For a wetware compiler that gets only very limited support from current tooling. I can't do that. Not even close. I need a language and tooling that's on the same page as me, to serve as an extended self. And for this, that's not Haskell.
Many years back, someone in a conversation suggested a core skill of a professional C++ programmer, was cataloging the things which seem like good ideas, and might even serve well in some other OO language, but which tended to go badly in that era's C++. Along with many baroque workaround to salvage others.
Fewer years back, I'd a conversation about using Haskell in education. We talked of alternate preludes to remove historical clutter, and extensions to remove historical limitations. And turned up a fun divergence - he thought of learning Haskell as fairly easy, and I thought of it as ghastly hard. For his default context was programming in the small - vocabulary, functional mindset and such. And mine was... you the start at the other end, with the dream, the no-extraneous-complexity pure essence of what is needed... and now you have to learn the zoo of baroque work arounds, and crippling limitations, by which Haskell fails the dream.