AI can pursue a goal. But who decides when it has gone far enough? On Frankenstein, AI misalignment, and why “enough” is still a human judgment.