FREN
Parentalité

AI isn't my kids' enemy — it's the most patient private tutor they'll ever have

28 August 2026 · 7 min

“Doomers gonna doom. Be the light.”

I came across that line and it hit home instantly. Because everything you read about AI and kids is fear: it’ll make them dumb, it’ll replace teachers, it’ll steal their future. Fear gets clicks. But it misses the point entirely.

Me, I’m a developer. I build AI agents all day long. And when I picture what it could change for my kids, what excites me most isn’t a product I shipped — it’s what it would change for the homeschooling we’re considering.

I have a daughter, Émilie, and a second child on the way. We’re seriously thinking about home education (IEF, as it’s known in France) for them. Naturally, I ask myself the question every parent asks: what kind of world will they grow up in, and how do I prepare them for it? My answer runs against the grain of the prevailing discourse.

The doom misses the best use

We hear about AI that “cheats for students.” We forget to mention what it could become for a parent who home-educates: the most patient, most available, most personalised private tutor a child could ever have.

A child could learn at their own pace. When a concept gets stuck, the AI doesn’t sigh, doesn’t lose patience, doesn’t make them feel they “should already get it.” It re-explains ten times, differently, with an example drawn from whatever they’re into. A human private tutor capable of that would cost a fortune and would never have that availability. That’s exactly what I’d want to be able to offer them, at home.

This isn’t the dystopia we’re being sold. It’s a gift — provided you know how to use it.

Why fear is a bad advisor

I understand where the worry comes from. We’ve watched digital tools turn childhood into a playground for capture algorithms: feeds that scroll forever, games engineered to hook, empty content churned out at scale. The wariness is healthy. But it’s aimed at the wrong target.

The problem was never the tool — it’s the intent behind the tool. A social network that maximises time-on-screen doesn’t share a goal with a tutor you’ve framed yourself to make a child think. Conflating the two because they both run on screens means throwing away the most powerful lever we’ve ever had, just because it looks, from a distance, like the thing that scares us.

Fear is also paralysing. It pushes you to ban rather than to teach. And banning prepares no one: it simply pushes the encounter with the tool to a moment when the child will be alone, with no compass, facing uses they were never taught to weigh. I’d far rather be next to my children as they learn to wield AI than turn it into forbidden fruit they’d discover in secret.

Concretely, how I’d go about it (and what I’d build)

Because I’m a developer, I wouldn’t settle for just “using ChatGPT.” If we take the leap into home education, I’d build custom tools for the day-to-day I picture:

  • Preparing lessons from real material. I’d photograph a textbook page, and an agent would produce a structured lesson plan, exercises tuned to the child’s level, and even illustrations on the theme that grabs them. The prep work that used to eat entire evenings would take a few minutes.
  • A calibrated tutor, not a parrot. The point wouldn’t be for the AI to give the answers, but to make them think. I’d frame it to ask questions, guide, and refuse to chew the work for them. Learning to learn, not to copy-paste.
  • “Slop-free” kids’ tools. The kids’ web is an ocean of mass-generated content — noisy and empty. Since I can code, I could build small in-house apps — clean, ad-free, with no attention manipulation — shaped exactly for what we’d want to work on.
  • Freeing creativity, not replacing it. AI would be there to amplify what the child wants to create (a story, a project, a model), not to do it for them. The hand would stay theirs.

Nothing magical here, and I’m not going to spin you tales of miraculous results: we’d learn, we’d adjust, it would be family build-in-public. But the direction is clear.

What strikes me, just imagining these tools, is how much AI would shift the home-educating parent’s work. Most of the energy usually goes into preparation: finding the right exercise, adapting an explanation, making a worksheet. That share would shrink, and the freed-up time would go back where it truly counts — next to the child, watching what clicks, digging into a question, riding their curiosity. The tool wouldn’t replace the parent; it would hand the parent back time to do what no machine ever will.

The guardrails (and this is where my job helps most)

Precisely because I know what’s behind the curtain — I know how these models hallucinate, how apps capture attention, how “slop” gets manufactured — I’d probably be more demanding than the average parent:

  1. Human first, always. AI is a tool inside the relationship, never a substitute. The bond, the debate, the “why,” that would stay between us.
  2. The screen stays a means, not a babysitter. AI would speed up prep so we spend more quality time together, not less.
  3. Always verify. An AI gets things wrong with total confidence. I’d teach the kids exactly to doubt, cross-check, and not swallow everything — a skill that’ll be worth gold in their world.
  4. Zero slop, zero manipulation. No app that exploits attention. If it doesn’t exist cleanly, I’d build it.

The real point: preparing them to wield AI, not fear it

My children will grow up in a world where AI is as ordinary as electricity. Making them afraid of it doesn’t protect them — it disarms them. The service I can do them is to teach them early to use it with discernment: as a powerful tool you steer, question, and stay in charge of.

There’s one skill that’s about to become central, and that almost nobody teaches: knowing how to converse with an AI. Stating clearly what you want, spotting when an answer rings false, asking again differently, cross-checking. It’s a blend of critical thinking, precision and curiosity — exactly what homeschooling could cultivate, without ever naming it. My children won’t become dependent on a machine; I want them to learn to use it like an instrument, keeping their hand on the wheel.

That’s also why I share this reflection. Not to lecture anyone — to show there’s another story to choose. While the doomers doom, each of us, at our own scale, can be the light.


Sébastien Debollivier — developer, solopreneur dad, I build products and AI agents. If you want to chat about AI at home (or want me to help you build a tool), drop me a line.

Questions fréquentes

Can AI really act as a tutor for homeschooling?

Framed well, it could. Set up properly, it would re-explain a concept as many times as needed, without getting impatient, using examples drawn from what the child is passionate about. The parent's role would stay central: AI would speed up prep and personalise explanations, but it's the human bond that carries the learning.

Won't AI just do the work for the child?

That's the real trap, and it's solved at the level of the instructions. I would frame the tool so it asks questions, guides, and refuses to hand over ready-made answers. The goal would be learning how to learn, not copy-pasting. It's exactly the kind of guardrail a parent who builds their own tools can enforce.

How do you avoid mass-produced kids' content (the slop)?

A huge share of the kids' web today is content churned out at scale — noisy and empty. When you know how to code, you can build small in-house apps: clean, ad-free, with no attention-grabbing tricks, shaped exactly for what you'd want to work on at home.

Do you need to be a developer to use AI for home education?

No. Most of the uses described here are within reach with a well-configured consumer tool. Knowing how to code lets you go further (custom tools, fine-grained guardrails), but the essentials — a patient tutor, faster prep — are available to any curious parent.

What guardrails should you put in place with your kids?

Four anchors I'd set for myself: human first (AI is a tool inside the relationship, never a replacement); the screen stays a means, not a babysitter; always verify, because AI gets things wrong with confidence; and zero slop, zero attention manipulation.

We're expecting our little boy in August 2026.

See the baby registry →
❤ La lettre De Bollivier

Famille, reconversion et projets en cours — un email de temps en temps, vrai et sans filtre.