I tell my database students that I don’t care if they use AI.

Some of them look at me like it’s a trick. It isn’t. AI is being used everywhere right now, and pretending it doesn’t exist in my classroom wouldn’t prepare them for anything. So I encourage it.

What I care about is whether they understand what they get back.

An LLM can write a SQL query faster than any of my students can. Faster than me, honestly. That part isn’t impressive anymore. What matters is whether the student can look at that query and tell me what it does, why it’s written that way, and whether it’s actually right. A query can run without a single error and still return the wrong answer. If you can’t read it, you won’t know.

I wrote recently about an AI-generated ADR that nobody read before it landed in my review queue. This is the classroom version of the same problem. If you paste in an answer you don’t understand, you haven’t learned anything. You’ve just pushed the problem to the moment someone asks you about it.

The way I think about this comes from how I use AI myself.

If I’m picking up a new tool or language, or I need to dig into something I don’t know well, I don’t just ask for the answer. I give it context: what I’m working on, the scenario, what I’m trying to get done. Then I pick its brain. What are the pros and cons of this approach? What would you do instead? Show me an example. Now show me another one. By the end I’m not just holding an answer. I understand why it’s the answer.

For a lot of us who’ve been doing this a while, that has replaced Stack Overflow. I remember what it used to be like. You’d search, land on a Medium article that covered half of what you needed, and then Medium would tell you you’d hit your free article limit. So you’d open an incognito window and keep going. Or you’d finally find a Stack Overflow answer to your exact question, except it was four years old. You’d try it anyway and pray it worked, or at least gave you a better error than the one you already had. That was learning. It just took forever.

Now I can have that conversation directly, and it’s about my problem instead of someone else’s.

That’s what I want my students doing. Not “write me a query that does X,” but “I’m stuck, help me figure this out.” You can set the AI up so it doesn’t hand you the answer right away. Tell it you’re learning SQL, that you want hints instead of solutions, and that it should explain what’s needed and let you take a shot at writing it first. I’ve shown my students a few prompts that do exactly that, along with the kinds of questions they can ask. Used that way, it’s another guide. A tutor that’s around at 2am the night before something is due.

The class is set up to find out whether that’s happening.

There’s a midterm and there are weekly quizzes from lecture, and those make sure the concepts are landing. But the real test is the final project. Every student comes up with a business and builds a working database for it. Then they present it to me and the rest of the class.

That presentation is where I find out what they actually learned. I get to ask questions. Why is this its own table? What happens to these rows if a customer gets deleted? Walk me through this query. If they built it by understanding each piece, those answers come easily, even if AI helped write every line. If they pasted their way through it, that shows up pretty quickly.

I’m not trying to catch anyone. I’m trying to get them ready for something that happens at every job: someone senior points at your work and asks why. “The AI wrote it” isn’t an answer there either.