If I read the docs correctly, you're literally supposed to send a description of what the UI should look like (which is the same as HTML + CSS: declarative code). How is this any different from just sending HTML + CSS?
Kind of reminds me of Mozilla XUL (rest in peace...)