AI-Era Apprenticeship

Rebuilding the training mechanism that entry-level work used to provide — real projects, real external judgment, short mentorship chains.

The thing we keep noticing

A lot of parents we know are feeling lost right now. Not loudly — more like a quiet unease. They sense the world their kids are growing up into is changing fast, but when it comes to actually deciding what to do differently, most of us default back to the same track we grew up with: study hard, get the credential, get the job. Not because we believe it's still the right answer, but because we don't have a clear alternative to point to.

We don't think this is a failure of imagination. We think it's a structural gap, and it's worth naming precisely before trying to fix it.

The mechanism, as best we can tell

For a long time, there was a simple economic deal underneath entry-level jobs: junior employees were cheap enough, relative to their output, that companies could afford to also be training them. The gap between what a junior produced and what they were paid functioned as an informal training subsidy. Over a few years, that junior became productive enough to be worth their cost — and the company recouped the investment.

AI is breaking that deal — not because junior work suddenly stopped mattering, but because the same output can now often be produced by a senior person working with AI tools, at lower cost than hiring and training a junior. Juniors aren't competing against an absolute bar anymore. They're competing against "senior + AI." That's a much harder bar to clear, and it changes the economics of hiring junior people at all.

This matters beyond any one industry, because the entry-level rung wasn't just a job — it was where a lot of tacit, hard-to-name skills used to get absorbed: how to scope an ambiguous problem, how to tell when your own judgment is good enough, when to ask for help, how to defend a decision to someone whose opinion has real stakes for you. None of that was written down anywhere. It got passed on by people, in the course of real work, under real supervision.

When that rung thins out, that transfer mechanism thins out with it — even for people who still get the job title.

Why we don't think this fixes itself

It would be reassuring if this were just a temporary dip that the market corrects on its own. We don't think it is, for a fairly mundane reason: training juniors has always been something closer to a collective good than an individual firm's optimal choice. Every company benefits from a pipeline of trained people existing somewhere — but no individual company is rewarded for being the one that pays to train them, especially when a trained person can leave for a competitor, or when AI now makes the investment less necessary for the marginal task. That free-rider problem existed before AI; AI just removed the thing that used to make the math work out anyway.

Government subsidies, corporate training mandates, and school curriculum reform are all reasonable responses to this kind of problem, and we support efforts at all three. But they're slow, by nature — they require institutions to move, and the institutions in question (large firms, legislatures, school systems) are exactly the slowest-moving parts of this whole system. We don't think the gap can wait for them alone.

What we're building instead

A small, deliberately un-fancy idea: rebuild the training mechanism of apprenticeship directly, without needing an employer's hiring economics to fund it.

The structure is simple. Someone earlier in their formation works on a real project — something with actual external users, not a practice exercise — under the loose guidance of someone who's only a step or two ahead, not necessarily a career ahead. The goal isn't transferring technical skill (AI is, honestly, a pretty good tutor for that part now). The goal is training a specific, nameable set of judgment moves that AI doesn't train on its own: scoping an unclear problem, verifying AI output against reality rather than plausibility, knowing how much scrutiny a decision deserves, choosing between approaches rather than just using the one in front of you, knowing when to escalate rather than guess, and being able to defend a judgment call after the fact to someone whose opinion actually matters.

When a cycle finishes, the person who went through it is qualified to mentor the next one — not because they've spent decades in the field, but because the edge needed here is being a little ahead, not far ahead. That's the part that makes this scalable in a way the old apprenticeship model, with its dependence on scarce senior mentors, never quite was.

What this is not trying to be

It's not a replacement for school, and it's not a credential. It's not trying to out-compete universities or bootcamps at the thing they already do reasonably well. It's specifically aimed at the part of formation that depended on a real job existing to deliver it, and that's becoming less reliable to count on.

We're starting in one domain — AI-assisted software development — partly because it's where we have real material and credibility to offer, and partly because it's one of the clearest places this disruption is already visible. But we don't think the underlying problem is software-specific, and we'd be glad to see this pattern tried in other fields by people closer to them than we are.