Given a point on a torus there exist exactly 4 circumferences that lie on the surface of the torus passing through that point. The two non-obvious ones are called Villarceau circles.
Wikipedia gives a proof of their existence with analytic geometry. I'm interested in a proof with synthetic geometry because of this anecdote by Emilio Segrè:
On a different occasion, while I was waiting to be called to an oral examination, Majorana gave me a synthetic proof for the existence of Villarceau's circles on a torus. I did not fully understand it, but memorized it on the spot. As I entered the examination room, Professor Pittarelli asked me, as was his wont, whether I had prepared a special topic. "Yes, on Villarceau's circles," I said, and I proceeded immediately to repeat Majorana's words before I forgot them. The professor was impressed and congratulated me on such an elegant proof, which was new to him.
(From here, emphasis mine)