Math tends to bug me because "proving" theorems seems to come from convincing arguments, but we have enough computation power at this point that we should reduce them to mechanical/machine verification and then just run them and know.

With enough rigor, you don't need to "convince" anyone--a machine can verify it.