For creating improved enzymes by directed evolution, it is thought that the use of larger libraries will yield variants with the most desirable properties. However, rapidly sifting through tens of thousands of variants to find that special enzyme can create a bottleneck. In this issue, Olsen et al. use FACS to rapidly sort through a large library of OmpT protease variants displayed on the surface of bacteria. This allowed them access to the substrate—a peptide containing a nonpreferred OmpT cleavage site attached to fluorescent FRET partners. Cleavage disrupts FRET quenching, yielding a fluorescent signal that corresponds to the activity of the protease variant. Using the method, they screened a large library of ∼6 × 105 random OmpT variants and found ones with 60-fold higher activities for the nonpreferred cleavage site. The method can potentially be applied to a variety of proteins displayed on the surface of bacteria or yeast (see p. 1071).