White has no good discoveries to unveil at the moment; his bishop on g2 is behind his knight, but the bishop lacks a good target. So White experiments with other moves and finds that two—Nc5 and Nd6—are of special interest: they both move the knight into the path of White’s other bishop on a3, creating the kernel of a discovery. Examine such moves carefully to see whether they might also force an enemy piece into place as a target of one of the pieces in the kernel. Nc5 is unattractive because it doesn’t force anything (it’s also bad because it lets Black uncramp his position by exchanging knights). But Nd6 is another matter; it attacks Black’s rook. Taking the knight with QxN is out for Black because the knight has protection; the rook must be moved instead. But where? It has three possible squares: g8 (leading to the smothered mate Nf7#), f8, and e7. The key point about either of the last two possibilities is that they move the rook onto the same diagonal as White’s bishop on a3, creating a target for it once it is unmasked. The knight then on d6 just needs a violent move to make as it unmasks the bishop. Giving check on f7 wouldn’t work because it enables the rook to both extinguish the threat and take itself out of trouble. The correct move is NxBc8. If Black recaptures, BxR wins the exchange for White on the next move; if Black doesn’t recapture and spends his move saving the rook, White has won a piece.
As this case shows, moves that create the kernel of a discovery are most interesting when they make a side threat against the enemy at the same time; for then he has to spend a move dealing with the threat rather than dealing with the discovered attack you've just constructed.