Do you ever analyze your games with a computer and find really unusual hidden possibilities? It can be worth it sometimes!
I came across this curious position while analysing one of my games from a recent tournament. Black has the initiative, but in most lines White is active enough to trade off all the pawns. The d2 pawn is a strength and a weakness at the same time. Computer found a brilliant concept:
53… g3!! You can check yourself that taking on f2, or f5 is not quite sufficient.
The idea is to win control over the e3 square, so after 54. fxg3 Re3 + Black wins the rook.