In the canonical mode of protein:DNA recognition, amino acids make specific contacts to nucleotides of a native DNA double helix. As RNA-guided gene editors and bacterial immune sentinels, CRISPR-Cas nucleases (clustered regularly interspaced short palindromic repeats, CRISPR-associated) break this mold, identifying target sequences instead through the formation of an ~20-base-pair R-loop. Requiring dramatic disruption of the DNA helix, this unique recognition mode is the basis for CRISPR’s powerful applications in research, medicine, and agriculture, but its physical underpinnings are poorly understood. In my thesis work, I studied the mechanisms by which Cas9 and Cas12 “sculpt” DNA to achieve their programmable solution to sequence-specific DNA binding and cutting. In Chapter 2, I describe my discovery that Cas9 bends and twists DNA to interrogate its sequence, a search mechanism that is likely shared by all DNA-targeting Cas enzymes. These DNA-bending actions, repeated over and over, constitute the slowest phase of genome editing, urging research into the rules that translate bending dynamics into target-capture rate. In Chapter 3, I present a yeast-display system designed to direct the evolution of fast-searching Cas mutants, both for basic inquiry into bend-assisted target search and for direct technological use. Finally, in contrast to Cas9, which forms double-strand breaks using two distinct nuclease active sites, Cas12 achieves the same feat using a single active site, necessitating further manipulation of DNA even after successful target capture. In Chapters 4-5, I show that Cas12 induces conformational plasticity in its substrate by exploiting a strand exchange junction with special structural properties, reshaping DNA in a way that maximizes the utility of minimal enzymatic machinery. While popular descriptions of CRISPR generally highlight its power as a pair of molecular scissors, my work demonstrates that the true functional novelty of these enzymes emerges from their skill as molecular sculptors.