Did I Actually Learn Anything?

It's pretty easy to ask AI to build an application now.

You describe an idea, wait for it to generate some code, fix a few errors with another prompt, and eventually something starts working.

I've seen a lot of people around me build things this way. But something about what happens afterward bothers me.

They don't really look at the architecture. They don't question the decisions AI made. They don't check whether there are problems beyond the errors that show up on the screen.

Then they share the project and say, "I made this."

And I keep wondering how much of what they made they actually understand.

I've been using AI too

I use AI a lot, especially while working on Recast and Zubience.

It helps me understand things, explore solutions, and get through work that would otherwise take much longer. I'm glad I have access to it.

But I didn't just describe an app and leave everything else to AI.

I planned how I wanted the projects to work. I thought about the architecture, how the different parts should fit together, and what each part needed to do. Then I explained those decisions to AI and asked it to help implement them.

I wasn't asking it to make every decision for me.

For the parts I already knew how to build, that saved me a lot of time. For the parts I didn't understand, I used it to ask questions and work toward understanding them.

Both were useful. But they required me to stay involved.

Working doesn't mean understood

An application opening without crashing doesn't tell me much about how well I understand it.

Could I explain what happens when someone clicks a button? Do I know where their data goes? If something breaks, do I have any idea where to start looking?

Those are the questions I think matter.

It's tempting to skip them when everything seems to work. Especially when I could spend that time adding another feature instead.

But if my only response to every problem is to copy the error into a chat and hope the next answer fixes it, there's a lot about my own project that I still don't understand.

That doesn't mean I need to memorize every line of code. I do need to understand enough to question what I'm accepting.

And planning the architecture myself doesn't automatically make the implementation correct. I still have to read what AI produces, check whether it follows the plan, and test the result.

Someone still has to take responsibility

This becomes more serious when an application handles other people's accounts, private files, or sensitive information.

A mistake can affect someone beyond the person who generated the code.

I wouldn't feel comfortable handing all of that responsibility to AI and assuming everything was fine because the application appeared to work.

If I publish something for other people to use, checking it is part of my job.

"AI wrote that part" wouldn't undo the damage if something went wrong.

That applies to my projects too.

So, did I actually learn anything?

I think the useful question is what I can understand and do after using AI.

Can I explain a decision I couldn't explain before? Can I spot a problem in its suggestion? Can I change something without asking it to rebuild the entire feature?

If I can, then it helped me learn something.

If I keep accepting answers I don't understand, I might finish the project faster while leaving myself just as confused as when I started.

I want to keep using AI. It saves me time, and being able to ask questions while building something is useful.

But I also want to be able to open my own projects and understand what I'm looking at.

When I say, "I made this," I want to be able to explain how it works.