No, I completely disagree. You need to design accessibility in from the start, it's almost impossible to retrofit. Very few systems manage to add high-quality accessibility later on.
I'm one of the Typst devs and I do agree with you here. LaTeX has a lot of trouble with accessibility because it's hard to retain semantic information through layers of macros. However, I think we are in a better starting position because Typst is designed to revolve around semantic elements that the compiler can actually understand. We haven't gotten to it yet (there's lots to do), but we want to use this information both to output Tagged PDFs and for semantic HTML export. I guess we'll see how it turns out!
I wish you the best of luck. I don’t have any time to get involved in any more open source projects, but I consider the lack of common accessible publishing formats for science one of the biggest embarrassments of academia — for a field that claims to be open, we sure seem to love churning horridly inaccessible PDFs (and yes, I’m as guilty as anyone else here).