Well, at least in math there's a somewhat formal notation for most things at this point? Obviously, mathematicians don't all agree on the precise notation, but there's generalizable ideas like "grouping symbols" (parens, brackets) and "left-grouping symbols" (upper-sigma, upper-pi), etc. etc. that inform the layout? Am I hopelessly misinformed on this? This kind of thing should suffice for 95%+ of things, and if you are an exception then you get to play in "lay it out yourself" land.

One issue that I had with LaTeX formulas was exactly the thing that parent poster was pointing out: You really had to describe the layout of the formula; not just its expression ("... and have at it, layout engine"). It is/was really just a kludge around the box model of TeX in the worst possible way.

Yes, you could produce the most beautiful formulas of them all, but that's just because you spent a huge amount of time learning (La)TeX and tweaking the exact expression of each individual formula[1]. Don't get me wrong, it's was still miles better than the competition at the time if you cared about the quality of the output -- not to mention the severe resource constrains -- but surely we can do better these days? Surely?!?

Btw, I don't think parent poster was talking "semantic web" at all, as I read it at least.

[1] Same deal with pseudocode. Oh god, the amount of time I spent tweaking pseodocode in my university years.