I stopped asking AI for answers. I asked it to quiz me instead. (Part 2)

Last week I shared how an AI-coached challenge helped me relearn Docker multi-stage builds in 20 minutes. This week, before we’d move on, we had a quick refresh on the previous lesson.

I took note of the named FROM, the package copy trick, the different images. Good. And off the cliff again.

Challenge 2: Docker networking. The AI gave me a broken setup: two containers that couldn’t talk to each other. I had no idea of the syntax. But I knew the concept — they need a shared network — so I just tried what made most sense.

Turns out, Docker is pretty good at making things as simple as possible. To my surprise, my syntax was actually correct. I’d just forgotten to create the network first. Three commands total, and it all made perfect sense. I tried to lock them in for the next session.

We had time left, so the AI switched to a quick-fire round: “What does a Docker setup look like with other languages?” Python, Java, PHP, Go… Each setup tells something about the language, not just about Docker — which needs a build step, which doesn’t, which drags along a heavy package manager you want to get rid of. I started building a mental map of how different stacks work under the hood.

Then we jumped straight into practice. We started with the simplest one, Python. By the time we got to PHP, time ran out. I decided to sleep on that exercise. The muscle memory was building.

Twenty minutes. A networking challenge, a syntax drill across four languages, and homework for tomorrow. I sat down to practice Docker — and walked out with a broader picture.

That’s one thing I enjoy about this format — the side-learning keeps things fun. You pull on one thread and something else comes along for free.