Am I learning to code or just learning how to ask AI for code?

I am still fairly new to building software, and AI has helped me finish things that would have taken me much longer on my own. But recently I noticed something that bothered me. I was building a small API route that creates a project and saves it to a database. I asked an AI coding tool to generate the route, validate the request, check the user, and insert the record. The code looked clean. The types looked correct. It even worked on the first few tests. Then I changed one field in the database and everything started failing. The error mentioned a transaction, the response returned the wrong status code, and one value was becoming null even though I thought it was required. I kept asking the AI to fix each error. Every answer added more code, but I understood less after every change. Eventually I realized that I could not explain the full request flow. I knew the request reached the API route. I knew some validation happened. I knew the database received something. But I could not clearly explain what happened between those steps or why the fix worked. So I tried the same idea again with a smaller route. This time I only used AI when I was stuck. I wrote the validation myself, logged the data at each step, and read about how the database client handled errors. It took much longer, but I could actually explain the result. Now I am unsure how to measure progress. With AI, I can finish more features. Without heavy AI use, I finish fewer things but understand them better. Both seem useful, but they are not the same kind of progress. Maybe the real skill is learning when to ask AI for code and when to struggle through the problem yourself. For people who use AI while learning development, how do you stop it from doing too much of the thinking? Do you have any rules for when AI is allowed to write code and when you force yourself to work it out?