I wrote a toy browser in a weekend too; I just used an existing HTML and CSS parser. The hard part isn't the HTML parsing, that's pretty easy and mostly just basic lexing/parsing stuff. It's applying the CSS to the HTML and rendering a document that vaguely looks right. There are loads of different interactions, and even Firefox and Chrome actually get it wrong in many edge cases (and also frequently in not-so-edge cases). I wrote a list some time ago of all the interactions "position: sticky" breaks in Firefox, and there's loads of them (I can't find that list now though, it was a HN comment from last year). It was pretty basic stuff like z-index, tables, flexbox, etc. Nothing too obscure.

If a lot of the cruft would be removed then actually, it probably wouldn't be so bad. But you know, compatibility and such. Still, a <link rel="newstylesheet" href="style.newcss"> probably wouldn't be such a bad idea. The base CSS specification can probably be reduced by half if not more, both by removing cruft and by learning from mistakes.