The pawn takes f5 and h5, and after Bg2 the bishop hits d5 through the hole the pawn left. The theory that exists is sharp and short: 1…d5 2.Bg2 Bxg4 3.c4! is the point, hitting d5 while the bishop is off the queenside.
It does not lose by force, but it is not free either: the engine already rates Black about a pawn better after 1…d5, the squares round White's king are permanently soft, and the opening's value is surprise. Played twice against the same opponent, it is a weakened kingside and nothing else.