Traditional, static documents will never go extinct, particularly in fields such as pure mathematics. If LaTeX3 wants to engage this crowd, it'd better offer substantial benefits, a fair amount of backward compatibility, or preferably both. In any event, LaTeX moved on from the original set of macros in 1983 to LaTeX 2.09 (anybody know when?) to LaTeX 2e in 1994... in the intervening quarter of a century LaTeX3 has been a work in progress... Maybe the goals set for the next version are too ambitious (see [1]), but clearly the inertia has been lost, a sign that the project has become fairly stable.
As with any long-term speculation, many years from now people may very well get rid of the TeX ecosystem altogether. During our lifetimes, though, I'm skeptical of seeing major changes in (pure) mathematics publishing. As unixhero pointed out, jupyter notebooks and the like will gain further adoption, and the applied math crowd, scientists and engineers will occasionally rant about having to choose between notebooks and LaTeX... c'est la vie!
As with any long-term speculation, many years from now people may very well get rid of the TeX ecosystem altogether. During our lifetimes, though, I'm skeptical of seeing major changes in (pure) mathematics publishing. As unixhero pointed out, jupyter notebooks and the like will gain further adoption, and the applied math crowd, scientists and engineers will occasionally rant about having to choose between notebooks and LaTeX... c'est la vie!
[1] https://tex.stackexchange.com/a/986/43202