[flohofwoe]

This is all trivial (in fact much, much easier) to do with a handful of tiny custom APIs instead of what HTML5 provides (there are plenty of native window system glue library examples which have much saner APIs for window and input handling: GLFW, a subset of SDL, winit...).

Rendering an UI that even remotely looks "native" is harder with HTML+CSS than a handful of straightforward rendering primitives which can be implemented on top of WebGPU.

High quality text rendering is a bit non-trivial of course, but that's also only an issue if you want to do a native-looking UI (but mostly irrelevant for games).