I remember reading about a chess puzzle where a similar "out of the box" thinking was required which would probably put it outside of reach of computer solvers. As a plus it was within currently legal rules.
The winning move was en passant and the solution required observing that the only way the situation could have arisen was the opposite pawn moving 2 ranks forward, making this move possible.
This doesn't seems like the kind of move that would be the crux of a puzzle, tripping up both computer and human. En passant captures are a normal, straightforward move in chess that few above amateur level would overlook. En passant captures are the first move you tend to look at if an opponent has moved their pawn two spaces adjacent to one of your own.
> En passant captures are the first move you tend to look at if an opponent has moved their pawn two spaces..
GP is clearly saying that you're not given the information that a pawn has moved two spaces, the trick of the puzzle is to infer from the board's layout that such must be the case.
Seems damned clever to me. From a quick google it appears that such things are known as "retrograde puzzles".
The winning move was en passant and the solution required observing that the only way the situation could have arisen was the opposite pawn moving 2 ranks forward, making this move possible.