I think this may be the source of many religious wars in frontend land. HTML was originally a document model, but now it does dual duty as a way to block off rectangular sections of application UI.
The purist approach of writing perfectly semantic HTML and placing all presentation in optional, layered style sheets works beautifully - for documents. But I don’t think it works as well when you are designing an interactive UI and you’re really just trying to manipulate pixels on a screen.
Tailwind sure is ugly, but it’s quick to write and a cinch to debug. And in my experience, most UI code is pretty ugly. It seems to come with the territory?