For me, it would just be more work without any benefit. For me, a pencil and paper is necessary. I can't think typing equations out. Part of it is probably just habit, but a good part of it is that pencil and paper support non-linear workflows effortlessly (e.g., drawing a box around a side calculation). In the end, I would end up writing the problem set up on paper, then transcribing it into TeXmacs or whatever for no real benefit. This isn't just theoretical. All through gradschool, I worked out research on paper before writing it up in latex for a paper or thesis section.