More practice questions won't help my child

29 July 2026 · by Larry

The other day I tried a small experiment. I asked an AI to write me twenty Primary 3 maths questions. It did it in about two seconds, and they were perfectly fine.

That moment turned a worry I'd been carrying into a clear thought: the questions themselves are not worth much any more. Anyone can make a thousand of them now, for almost nothing. Plenty of people could build a tool that drills a child on endless sums — the AI does the hard part for free.

So if the questions are the cheap bit, what is the part that actually matters?

The part a pile of questions can't give you

Think about a good teacher, or a patient parent at the kitchen table. They don't just hand over more questions. They notice. They see that your child gets subtraction right until there's a zero to borrow across, and then it falls apart. They catch the one thing that got a little better this week, and say so, even when the score is still not great.

That noticing is the expensive part — and it's exactly what a stack of extra worksheets gives you none of. A child can do a hundred more sums and still nobody has told them what, precisely, they keep getting wrong, or that they're improving at all. They just do them, get a number, and move on. So they play once or twice and drift away, and who could blame them.

What I tried to build in instead

I can't put a real teacher inside a web app, and I'm not going to pretend I have. But I did try to build in a little of the noticing.

In StudyLab, when a question offers wrong answers to choose from, those wrong answers are not random. Each one is a real mistake a child actually makes — the classic slip, the rule half-remembered. I wrote them that way on purpose. So when a child picks a wrong one, it isn't only "wrong". It quietly points at which idea hasn't clicked yet.

"Fix My Mistakes" then takes just the ones they got wrong and walks through why — in words, not only the right number. And the small things that keep a child coming back are there too: a streak, stars, and a little buddy that grows from an egg the more they practise. Not a score shouting at them — gentle nudges that say you're getting somewhere.

What I'm still building toward

Here's the honest bit. Right now the child sees all of this; the parent doesn't, not really. What I'm slowly working toward is a way for you to open it and see the one thing worth sitting down to help with today — not a wall of numbers, just something like "she keeps getting stuck borrowing across a zero." That doesn't fully exist yet. I'm building it carefully, because getting it wrong would be worse than not having it at all.

And I want to be careful about what a tool like this can and can't know. A wrong answer can hint at which idea tripped a child up. It can't tell you why they were off today — whether they were tired, or rushing, or just not in the mood. That part is still yours, as the parent. The most the app can ever do is hand you a better place to start the conversation.

AI is a wonderful helper for a child who's learning. But it's a helper — a tool. The growing, the being noticed, the small "well done" landing at the right moment: that's the thing I actually care about building. The questions were never the point.

— Larry