For most of computing history, one thing held true: a computer would only do what you—or someone—explicitly told it to. That principle shaped how developers debugged, how IT pros solved problems, and how organizations thought about value and control. But with the rise of LLMs and generative AI, that truism no longer holds. Inputs don’t guarantee consistent outputs. Models behave probabilistically. And companies now invest in technology that may—or may not—do what they want it to. This episode reflects on the unsettling, fascinating shift toward tools that feel intelligent but resist predictability—and why that might be a feature, not a flaw.