Engineering Craft › AI-Assisted Development

Learning While Using AI

Using AI without skipping the understanding a junior needs to build.

Also known as: learning with AI, AI and learning, over-reliance on AI

AI assistants can write working code in seconds, which makes it easy to ship something you don’t understand. When you’re starting out, the risk isn’t that the AI is wrong. It’s that you stop building the skills that let you notice.

The traps

• Pasting without reading. If you can’t explain a line, you can’t debug it, change it or defend it in review. It’s now your code.
• Skipping the struggle. Working through a problem is how understanding forms. If AI solves every problem first, you get the answer without the learning.
• Trusting confident answers. AI output can be wrong, outdated or invent functions that don’t exist (hallucination). It sounds equally sure either way.
• Fixing symptoms in a loop. Pasting each error back and accepting each patch can pile up changes that no one understands.
• Not knowing what to ask. Without fundamentals you can’t judge a good prompt or answer.

Habits that keep you learning

1. Try first, even briefly, so you have a mental model to compare against.
2. Ask for explanations, not only code: “why does this work?”, “what are the alternatives and trade-offs?”
3. Explain it back in your own words, or to a teammate. If you can’t, ask again.
4. Run it, test it and read the diff before keeping anything (reviewing AI code).
5. Check against the official docs when it matters (reading docs).
6. Keep practicing the basics by hand: reading errors, using the debugger, writing small functions from scratch.

A good test: could you rewrite this solution from memory tomorrow, and would you notice if it were subtly wrong? Used that way, AI is a patient tutor and a fast pair of hands. Used otherwise, it hides the gaps you most need to see. Also see vibe coding for what happens at the extreme.