Make the group difference depend on x

See how a main effect sets the baseline gap while an interaction changes the gap across the predictor.