Two Nature Biotechnology reports describe computationally guided routes to alter protospacer‑adjacent motif (PAM) recognition of CRISPR–Cas enzymes. One team developed Protein2PAM, a protein language model and evolution‑informed pipeline that predicted PAM specificity and produced Cas9 variants with broadened PAM compatibility and increased activity. A companion paper used protein language models to customize PAM recognition across Cas enzymes. Both studies combined in silico mutagenesis with laboratory validation to unlock genomic sites previously inaccessible due to PAM constraints. The work offers a practical path to expand targetable sequences for therapeutic genome editing and to accelerate engineering of bespoke CRISPR tools for specific genomic loci.