Researchers reported in Science that AI-guided protein design can produce functional synthetic RNA-guided nucleases with activity matching or exceeding natural enzymes. The work uses a hybrid of ESM Inverse Folding and evolution-informed constraints to expand protein sequence space while preserving the coordinated conformational states needed for RNA-guided cleavage. The authors engineered variants of TnpB, a CRISPR-Cas12-like family that couples programmable nucleic acid targeting with additional biological functions. In cell-based testing, the synthetic nucleases showed activity consistent with the designed strategy. The implication for genome engineering is straightforward: AI can generate non-natural nuclease architectures with potentially improved properties, which could broaden the CRISPR toolbox beyond sequence-closest variants.