“What is the best X for Y” — scored for cross-model agreement, not for which name wins. Ranked by cross-model divergence — the questions where the machines disagree most are first.