Two Nature Biotechnology papers reported computational and experimental platforms that use protein language models to predict and reprogram protospacer adjacent motif (PAM) specificity of CRISPR–Cas enzymes. One study presented a model (Protein2PAM) that predicts PAM specificity and guided in silico mutagenesis; the other used model‑guided evolution to generate Cas9 variants with broadened PAM compatibility and improved activity. The paired publications show how AI‑driven protein engineering can expand the genomic sites accessible to CRISPR editors, lowering a key practical barrier in therapeutic genome editing. Developers of gene‑editing therapies can use these methods to design nuclease variants tailored to clinical targets that were previously unreachable due to PAM constraints.