Originally published in UX Collective. This essay examines what happens when AI learns to validate instead of challenge, and why that seemingly harmless design choice carries consequences far more serious than a technical bug. Drawing on the GPT-4o sycophancy incident, the myth of Narcissus and Echo, and the distinction between passive and active use of AI, the piece traces the quiet erosion of intellectual friction in our most intimate cognitive spaces. Written from inside the experience: as someone who uses AI daily as a personal critic, and who almost got lost in its flattery before recognizing the trap. The central argument is simple but uncomfortable: a compass that always points in the direction you want to go has stopped being a compass. The value was never in the answer. It was in the resistance.