The apprenticeship problem is what happens when the most automatable work is also the work through which humans are formed. Junior coding, legal research, basic analysis, entry-level design, clerical reasoning — these look like obvious efficiency wins, but they are also where people acquire judgment, tacit knowledge, error experience, mentorship, and a professional identity. Automate them and near-term output can rise while the pipeline that produces future experts quietly closes.
It is a human-capacity transfer and a classic ecological externality: an individual firm has little incentive to preserve a society-wide training pipeline it does not own, so each locally rational decision to automate entry-level work can add up to a civilization that has stopped forming the seniors it will depend on. The governing question is long-range and simple — where will the experts of a generation from now come from, if the tasks that formed them were automated a generation earlier? — which is why the framework treats protected human competence as infrastructure rather than nostalgia.