- I think it's orders of magnitude easier, we have abandoned easy and semantically significative HTML years ago ~thanks~ to frameworks like React. I would say that fine-tuned LLMs are one feasible and fast approach to this problem.
- I'm using the subset of HTML useful for my scope, no more than 2-3 hours if you already know some HTML is enough.
- W3C is enough for a project of this size I suppose.