I feel the same joy when seeing a derivation of a novel algorithm where the effective procedure falls effortlessly out of the definitions. A good example is Dijkstra's derivation of Smoothsort[1]. It's worth noting that he was educated as a professional mathematician, not as a computing scientist.

[1] https://www.cs.utexas.edu/~EWD/transcriptions/EWD07xx/EWD796...