All yeast selection experiments were initiated by transforming selection strains with a defined quantity of gel-purified Cas9 plasmid fragments for in vivo assembly. For library selection experiments, 130 ng backbone DNA was co-delivered with PID library insert at 1:0.1 M ratios. Owing to the genetic readout inherent to SSA reporters and our use of a constitutive expression system, this molar ratio was critical for ensuring segregation of different library inserts into separate cells upon transformation, and thus reliable coupling between phenotype and genotype. The total number of transformants plated with positive selection (3.9e5 for the round 1 library and 4.5e4 for the round 3 library) or dual selection (3.9e5 for the round 1 library and 1.2e5 for the round 3 library) was estimated from the concentration of CFU quantified on SC-Leu-Ura media. Well-isolated single colonies were randomly chosen from selection plates for patching onto the same media and cell mass was subsequently picked from patches for sequencing by colony PCR. A proline codon was inadvertently duplicated external to the NheI restriction site during design of the pGG211 backbone used for Cas9 assembly in round 1, and this presumably resulted in the elevated frequency of Q1091P substitutions observed in those selections (Supplementary Table 3). This was corrected in the Cas9-Zif268 (pGG442) and isogenic Cas9 control (pGG431) backbones. For clonal selection experiments, 150 ng backbone DNA was co-delivered with PID library insert at 1:3 M ratios. Cells were serially diluted and plated in parallel on SC-Leu-Ura and positive-selection media (as well as dual-selection media, where applicable). After 48 h at 30 °C (or 72 h for dual selection), plates were photographed and then stored at benchtop for up to 16 h to facilitate scoring by eye. After this time the percentage of LYS2 + transformants was calculated as the ratio of LYS2 + , LEU2 + , URA3 + CFUs obtained with positive-selection media over the total transformant (LEU2 + , URA3 + ) CFUs obtained on SC-Leu-Ura. All library and clonal PID inserts were flanked by ~60 bp homology arms for assembly into backbones, and were generated with PCR or error-prone PCR using primers oGG629/oGG630. Backbone and insert fragment concentrations were quantified prior to experiments using the Qubit dsDNA Broad Range kit (Invitrogen) for library selections or a NanoDrop Spectrophotometer (Thermo) for clonal selections.