Because it's how normal people think. This under that. This after that, left-aligned with that. Text modes are the most basic tech supporting it.

Css uses boxing with no sugar (like size groups or custom allocation). Boxing is this in that in that in that, sometimes stretched. And now when you want A and B in different boxes to align well, you're screwed, cause it creates an invisible fragile tree of braces that hold everything in place.

Maybe they would use a much more natural constraints-based language instead of text modes. Like, literally, "this under that, and these five left- and width-aligned". But web is reluctant to add it, because when not used carefully it could slow down first interactive renders by 0.01% out of average 10s (that is, assuming the same moronic site structure for some reason and the code that renders it).