Directed evolution of stereo- and regioselective enzymes constitutes a prolific source of catalysts for asymmetric transformations in organic chemistry. In this endeavor (iterative) saturation mutagenesis at sites lining the binding pocket of enzymes has emerged as the method of choice, but uncertainties regarding the question of how to group many residues into randomization sites and how to choose optimal upward pathways persist. Two new approaches promise to beat the numbers problem effectively. One utilizes a single amino acid as building block for the randomization of a 10-residue site, the other also employs only one but possibly different amino acid at each position of a 9-residue site. The small but smart libraries provide highly enantioselective epoxide hydrolase or lipase mutants, respectively.