A coding agent already knows how to generate whatever it needs on the backend plus the client-side pieces. What's the point in having a Rust subset that compiles down to JS when I can just not have it and the result be the same?
I feel like the question these days is to figure out the right level of abstraction and Topcoat might be too much.
PS: This isn't a jab at the library or its authors. It's just me trying to figure out what still matters in the age of LLMs.