Math is all about being precise, logical.. Communicating exactly one concept at a time. Natural languages do neither.
Except math is almost never actually done that way in practice. Euclid was wonderful, but almost all modern math does not work that strictly (and Euclid really should have been more careful with the parallel postulate -- there's "more than one thing at a time" involved there). Yes, proofs are careful and detailed, but so is, say, technical writing in English. Except for a few cases (check out metamath.org, or Homotopy Type Theory) almost no-one actually pedantically lays out all the formal steps introducing "only one concept at a time".