Collaborate! And, if you dare, enter into a long term relationship. The vast majority of mathematicians I know do most of their work alone and, when they collaborate, the relationships are often short term temporary ones. But it can be extraordinarily fruitful to commit to a long term collaboration. I recommend it highly.
EDIT: Inserted "I know" after "vast majority of mathematicians" to account for Ben's comment, which I have no reason to doubt.

