Via John Baez: Mathematics for theoretical physics, a 700-odd page self-contained reference to all the maths you supposedly need to have some idea what contemporary theoretical physicists are talking about.
I saw this pop on arXiv, but I doubt it will be very useful to anyone who doesn't know most of the stuff in it beforehand. The exposition is pretty terse, and there are well over 1000 stated theorems.
If it's worth saying, but not worth its own post (even in Discussion), then it goes here.