I love typst. It's about time we have an authoring system like it that also let's you publish HTML versions of things you write with it instead of just PDFs. Papers in Academia for the longest time have been PDFs, which screen readers are terrible with, mainly because the PDF format is an abomination, and makes accessibility (let alone interoperability) unnecessarily painful.
PDF serves an important purpose; it’s not an abomination. It conveys exactly what the author or publisher intended. HTML can not do this. It can serve as a reliable archive and be made (imperfectly) accessible, as described in the article. Of course all this comes with significant tradeoffs, including reduced interoperability, as you say. But there are good reasons for academic publishing, especially in math, CS, and physics (but also in some areas of the humanities) to prefer PDFs—it’s not just obstinacy.