Shadowing, Not Pseudo-Apprenticeship
Watching an agent code teaches you only if you move past watching: question its choices, take over gradually, then work without it.
Contents
Use it when You are learning a new language or stack by watching an AI agent write code in it.
Pseudo-apprenticeship looks like apprenticeship but isn’t. You’re in the room with the expert, but you never pick up the tools. You watch the AI, accept the output, and stop there.
Shadowing is the deliberate version. Like pairing with a senior colleague who knows the stack, you push yourself through every stage:
- Watch the agent solve a real problem.
- Question its choices: why this pattern, what breaks if I do it differently?
- Take over partially: write the next function yourself and let the agent review it.
- Reflect: compare its approach to what you would have done.
- Work without it. That is where the real test happens.
It needs a foundation. It works when you already have a mental model for good code and map the new stack onto it. A true beginner is just watching someone type fast.
Slow the speed down. Read the output like a code review, line by line. Use the diff. Ask the agent to explain what it will do before it does it.
Watching passively is just television. Watching while questioning, then trying it yourself, is closer to real apprenticeship.