A study published in *Nature* describes how AI-redesigned starting points can enhance protein evolution. Directed evolution, particularly phage-assisted continuous evolution (PACE), can perform "dozens of generations of mutation, selection and replication every 24 h" to generate enzymes with new functions. However, evolution experiments struggle when mutations cause "excessive folding or solubility deficiencies," as "most WT proteins are marginally stable" and accumulating mutations often leads to instability. This trade-off can result in evolved enzymes having "lower activity or specificity for their new functions than their WT starting points." One existing approach to increase evolutionary potential is to first evolve an enzyme for "enhanced thermal stability or soluble expression," but this "requires substantial additional time and labour." The article body does not specify the exact AI method or outcomes, but the title indicates that AI-redesigned starting points and outcomes improve the protein evolution process.