Effort First, AI Second

Kids are using AI more and more, and honestly, who can blame them. The models keep getting better. They’ll work a physics problem, balance a chemistry equation, write the essay on the French Revolution, do the code, and then explain any of it back to you. For a student, it has never been easier to learn through AI.

That word, learn, is doing a lot of work in that sentence.

Because there’s a catch when you hand your thinking to a tool. The more you offload, the less your brain actually holds onto. There’s a name for it, cognitive offloading, and it’s pretty well studied. The work still gets done. The learning just doesn’t.

And none of this is going to stop. The cat’s out of the bag. Kids are kids, and kids take the easier path every time. You can call it a discipline problem if you want, but really it’s just human, and it might even be part of a real shift in how people learn. It’s not going back in the box. So the question I care about isn’t how to ban it. It’s how to make sure your kid comes out ahead anyway.

What the easy path actually looks like

Here’s how it usually goes. Kid gets home, feeds the homework to the AI, it spits out the answers, he submits it, boom, 100%. And if the test isn’t too far from the homework, he’ll probably ace that too, because it’s basically the same thing. So from the outside, everything looks fine.

But what’s actually getting built here is false confidence. The grade is real. What’s underneath it isn’t.

To see where that leads, follow three students with the same ability and the same ambition. The only thing different is how they use AI. Watch them from ninth grade through college engineering:

Three relationships to the same tool:

• The Offloader asks AI first and turns in the answer.
• The Abstainer won’t touch it. Pen, paper, textbook, YouTube.
• The Builder tries everything himself first, then uses AI to check, quiz, and explain.

The vertical axis isn’t homework grades. It’s whether the kid can solve a problem they’ve never seen before. And early on, the Offloader looks like the smartest one in the room. That’s the trap.

The part that makes it so hard to fight

Most of the warnings about this miss why it’s actually winning. Offloading doesn’t just buy you a grade. It buys you time.

The kid who offloads the homework gets his whole evening back. He plays more sports, sees his friends, games, sleeps, actually has a life. And to a teenager, that’s not a small thing. That free time might be most of what matters. And to be clear, the free time isn’t the bad guy here. Rest and friends and a life outside school are good for a kid.

Now look at the kid doing it the slow, honest way. She might keep up academically, and early on she does, but she’s paying for it in hours. Hours she doesn’t get to spend on the field or with friends. So the real cost of not using AI, at least at first, isn’t an academic one. It’s social. And that’s a rough thing to ask of a smart kid, to keep paying it while everyone around her coasts.

Once you’re in that spiral, once you’ve felt the advantage, it’s almost impossible to step back. Why would you? Then the other kids notice, and they start asking the obvious question: why am I grinding away when he isn’t? So more of them switch over, more of the thinking gets handed off, and the whole idea of desirable difficulty, the productive struggle that’s actually where understanding comes from, stops making sense to anyone. Why would you want things harder when the tool is sitting right there, and always will be?

And this is roughly where things already sit. Most kids are on the easy path, and the share keeps growing:

Everything hinges on the assessment

Whether any of this actually matters comes down to one thing. What do the tests actually test?

If they’re loose, if they just reward the same patterns the homework did, the Offloader sails right through. Good grades, good GPA, good school, maybe a good chunk of college too. Nothing ever checks the foundation, so the missing foundation never shows up.

But education is supposed to get harder. That’s the entire point of it, to keep pushing you into things you can’t do yet. And the further you go, the more every course assumes you actually learned the last one. A kid running on borrowed answers can fake the early stuff. What they can’t fake is the level where the material finally needs the understanding they skipped. And in something like engineering, where everything stacks on everything, that wall shows up early and it’s brutal.

Where this is heading

Play it forward and the pull just compounds. More kids offload, fewer struggle, and leaning on the tool slowly becomes the default. First for school, then for work.

Which gets at the question every parent is really worried about. In a world where a machine can spit out any answer in seconds, what makes a person worth hiring at all?

This is the part that actually gives me some hope, and the reason the slow path still matters. When the answer itself is free and instant, the valuable skill isn’t producing an answer anymore. Everybody can do that now. The skill becomes judging one. Knowing when the AI is confidently wrong. Asking it a sharper question. Steering it instead of just trusting it. That’s judgment, and you build it the exact slow way the Builder does, by grinding through enough real problems that you develop a feel for what right even looks like. The kid who only ever pasted in a prompt never builds that feel. So he can’t supervise the machine. He can only hope it got the answer right.

It’s the same reason we still teach kids arithmetic even though everyone has a calculator. Not so they can out-multiply the machine, but so they can tell when it hands back something that’s obviously nonsense. The struggle was never about beating the tool. It was about becoming someone who can actually run it.

The real problem isn’t AI. It’s motivation.

AI isn’t going anywhere, and here’s the thing, it can genuinely make a student better. The Builder is the proof. Effort first, then AI as a coach to pressure-test your thinking and explain what you missed. Do it in that order and you come out with the understanding and the fluency with the tool everything now runs on.

The hard part is getting a kid to actually choose that, because you’re asking them to think long-term, and most kids just aren’t wired that way yet. You’re asking them to spend more time, look slower, and feel behind their friends right now, in exchange for something they can’t see and won’t feel for years. Good luck selling that. Telling a fourteen-year-old to “embrace the struggle” is not going to work.

What does work is making the dead end visible. Letting the kid feel, early and concretely, that the easy route doesn’t actually hold up. A lot of that comes down to the structure around them. Assessments that really test understanding, so the gap between borrowed answers and real fluency shows up while there’s still time to fix it. Honestly, it’s less about a teenager’s willpower and more about whether the system bothers to check.

Now, you can’t go rewrite your kid’s exams. But the same lever works at home, just smaller. And it’s mostly about setting up a structure, not winning an argument:

1. Attempt first, always. Nothing goes into a chatbot until there’s a real, honest try on paper, even a wrong one. The effort earns the tool.
2. Let AI explain, not answer. “Show me where I went wrong” and “quiz me on this” build understanding. “Just give me the answer” takes it away.
3. Watch the homework-vs-test gap. It’s the earliest, clearest signal you’ll get. A’s on homework and C’s on the proctored test means the foundation is being borrowed, not built. And that’s a much kinder warning to get at fourteen than at twenty.

None of this means hovering over every session or learning the tools yourself. It just means making the easy path a little less automatic, and paying attention if that gap starts to open up.

The bet that pays off is the same one it always has been. Do the hard part yourself first, and let the tool sharpen it instead of replacing it. Get the order right and that early tax, the slower and messier and more frustrating start, pays itself back with interest. Get it backwards and your kid spends the later years paying off the effort they skipped, at a much worse rate.

It’s the same thing I keep landing on in What is the right way to learn?. The struggle isn’t in the way of the learning. The struggle is the learning.