AI is hollowing out junior engineers; preceptorship is the fix.

The core problem. Scott Hanselman argues that AI's ability to generate code is creating a crisis for early-career software engineers. Historically, junior developers learned by doing the "wax on, wax off" work—the tedious, routine coding tasks that built foundational skills. Now that AI handles those tasks, newcomers enter the profession at a higher altitude, skipping essential learning experiences. Hanselman warns this creates a hollowed-out pipeline where seniors retire without qualified replacements.

The preceptor model. Hanselman's proposed solution, co-developed with Mark Roussinovich, is borrowed from nursing: the formal preceptor. Unlike an ad-hoc tech lead mentoring an intern, a preceptor has an explicit, measured job: "mint new senior engineers." Their performance is evaluated not on code shipped but on engineers grown. Microsoft piloted this model with 400 people and found it successful, though Hanselman acknowledges it requires organizational willpower to prioritize long-term human development over short-term shipping velocity.

The organizational cost. The tension is economic. Preceptorship costs money upfront with delayed returns—happier, more productive engineers who stay longer. Hanselman contrasts this with the dominant approach: short-sighted quarterly optimization that poaches seniors from competitors. He warns that companies failing to grow their own talent will import competitors' culture and practices, losing organizational identity. The true threat isn't AI but "hyper-optimization by some suite executive" who treats engineers as interchangeable.

Remote work compounds the problem. Hanselman identifies a "perfect storm" of remote work, social media addiction, and AI that degrades junior development. He argues that human beings need in-person interaction, whiteboard sessions, and a "third place" beyond work and home. While he acknowledges remote work is possible with deliberate metacognition, he concludes it fundamentally makes mentorship harder, particularly for early-career engineers who need to shake hands and read body language.

The humanistic dimension. A striking part of the conversation is Hanselman's insistence that productivity gains cannot grow 3% infinitely. He calls infinite growth "by definition cancer" and asks what society does with AI-generated productivity—whether it becomes more leisure or more work. He cites studies about AI making engineers lonely and disrupting flow states, arguing teams must explicitly attend to the psychology and sociology of AI adoption, not just the technical benefits.

What's missing from the argument. The discussion remains mostly anecdotal and aspirational. Hanselman mentions a single Microsoft cohort but offers no detailed metrics on retention, skill acquisition rates, or cost-benefit analysis. There's no comparison with existing apprenticeship programs at companies like Google or Amazon, and no discussion of how preceptorship scales beyond 400-person pilots. The conversation also assumes a traditional office environment, which many companies have permanently abandoned.