I think we can have it all.
If we capture our various patterns and abstractions for the frontend in Haskell code, we can hide Tailwind from our frontend developers almost completely while making sure that our styles are consistent across our UI components....
The main benefit I derive from Tailwind is its discoverability. It seems easier to figure out how to do something with Tailwind than with plain CSS, and I think that’s mostly due to the information architecture of the docs and the library itself....