So, the large constants involve make the idea impractical, however theoretically interesting it might be. But it is possible to do exhaustive search of the space of programs in more restrictive domains; there's been some research, and there's a tool called "superopt" for finding optimal instruction sequences.

With Steemit there's no smart contracts, or even proof-of-work, that makes program optimization interesting on-blockchain. But it might be cool to think about what sort of "search" one might do over, say, all possible SteemSQL queries, if there was a worthwhile and efficient way of evaluating them.