Summer 2026 · A Mira Rethink
Rethink: Education & AI
Ready for the Tool
Preparing Children to Lead AI
Five rooms of educators, technologists, and parents stopped asking how children should use AI. They asked what has to be true about the person holding the tool. This is what they found.
What’s a Rethink?
A Rethink is a chance to think at full size about something we have stopped questioning. Most of what shapes a life is a social construct we inherited fully built: how children are raised, what a school is for, what work is, what care looks like. When one of those stops working, the instinct is to rearrange the parts — and rearranging is where thinking gets stuck. A Rethink goes back to first principles instead and asks what we would design if we were designing it now. It does not accept the question as given; it opens the question wider, until there is room to imagine something other than the version we already have. What comes out is not a solution and not a critique. It is a framework — enough structure that a new way of seeing a thing becomes something a person can actually begin to build on.
We believe seeing it differently is the first step to making it differently. This is that step.
Introduction
Every one of our conversations arrived, sooner or later, at the same crossroads: if AI can do more and more of what people do, what happens to people?
For a while we asked that question the scarcity way — what’s left for humans to do? — and it led nowhere useful. A machine that can write poetry does not make poetry one bit less worth writing; nothing you love doing becomes less yours because something else can also do it. The scarcity question turned out to be hiding the real one, which is not about the machine’s capabilities at all. It is about direction. Who decides what the tool is for?
AI is not a neutral tool. It is a force capable of reshaping the trajectory of a life, and of civilization, for better or worse. Followed carelessly, it flattens ambition, replaces judgment instead of sharpening it, and turns people into passengers of their own future. Led deliberately, it does the opposite: it takes the mechanical and the repetitive off our hands so that our finite time can go where we choose to spend it — imagining, connecting, creating, deciding what matters.
The difference between those two futures is not in the tool. It is in the person holding it.
The framework that follows is our answer to the crossroads. It is not a set of guardrails written out of fear. It is a design philosophy written out of conviction: that the measure of a good AI system is not how much it can do instead of us, but how much more it enables us to become. And the measure of a person using one — at nine or at forty-nine — is a single question this report will keep asking. Are you leading the tool, or is the tool leading you?
Where It Started
How do you prepare a child to use a table saw?
Think about what has to be true about the child. They have built things with simpler tools, so the saw has a purpose. They have made mistakes with cheaper equipment, so they know materials push back. Someone is standing next to them. They respect what the blade can do.
Almost none of that is about the saw. It is about the child.
Woodworkers also have a word for the moment the tool takes over. Kickback: the blade catches the workpiece and hurls it back at the person feeding it. Nearly everything in shop practice — the stance, the guard, the push stick, the habit of respect — exists for one purpose: to keep the person, and not the blade, deciding the cut. And nobody checks those things once and never again. You check them every cut. The most experienced sawyer in the shop is still checking.
That is the version of ready this report means. Not the day you are finally allowed to touch the saw, but what has to stay true every time you use it.
We asked the same thing about AI. Not how children should use it — every school district, dinner table, and boardroom is already arguing about that. Not when they should be allowed to — no birthday settles it. We asked what has to be true about the person holding the tool, so that the person leads the tool instead of being led by it. We started with children, because childhood is where those capacities get built — and where they are easiest to never build at all.
Children are using AI now. That will probably not change. What can change is whether anyone thinks about what has to be true first — and keeps checking — so the tool helps them grow instead of doing their growing for them.
Between April and July of 2026 we put the question to four rooms of people who take children seriously. Here is what they found.
The Framework
Back to the saw. Here is what has to be true about the person feeding it.
Six capacities came out of the rooms. They fit on a wall — short enough to remember, specific enough to use.
Read them as a gauge, not a gate. Nobody clears this once and is done — at nine or at forty-nine — because every capacity on the card fades quietly. The reflex to hand over the hard part arrives a little sooner each time. So check before you reach for the tool or offer it to a child. Then check after, and come back every few months.
Nobody administers it. No examiner, no certificate. A school that turns the six into an admissions test has broken them. The card is read by the person holding the tool.
The card is written in the second person on purpose. It means the same thing taped above a fourth-grade whiteboard or your office workspace. A child is simply where these capacities get built first — and where they are most breakable.
Ready for
the Tool
You’re leading the tool when you can…
- 01Sit with hard thingsWhat hard thing am I avoiding?
- 02Know your own voiceIs this mine? Am I proud of it?
- 03Ask what is trueWhere did this come from? Can I trust it?
- 04Collide & collaborateWho might see this differently?
- 05Teach what you knowCan I explain it without the tool?
- 06Add past the machineWhat did I add?
What This Looks Like on Monday
Here is where the six capacities touch the ground.
If you teach
Are you teaching kids how to ask these questions?
Before you assign something, check in with the framework. If your students choose to use AI, make sure that the learning process will be led by them and not AI. Use the questions to guide how you structure an assignment.
If you lead a school
Are you building students who will ask these questions?
Does your AI policy empower students to lead AI rather than be led by it? Are you building these six capacities throughout your organization? Create a culture where these capacities are valued and practiced and encourage staff to do the same.
If you are raising a child
Are you asking these questions?
Watch them use AI. Notice what they ask, what they accept, what they skip. Ask your child, in the car, at dinner: How has AI helped you lately? Where was it useless, or wrong? Ask what they trust AI with and what they don’t and why. Then let them teach you something they know well, and let them quiz the machine on it and find where it breaks.
Whoever you are
Run the card on yourself.
Before you hand the tool your next task, ask yourself, “Is there something hard here I am avoiding that would be worth working through?” After: “Is this something I would have created on my own, and am I proud to have it represent me? What did I add past what the tool created? Did I lead the interaction or was I being led? Is using AI making me and my work better and more effective? Or lazier and more disconnected?” If you are drifting toward the latter, what needs to change?
How This Was Made
Four sessions, ninety minutes each, between April and July of 2026. Six to ten people in each room: classroom teachers, professors, curriculum and district staff, technologists, movement-builders, parents. They called in from Virginia, California, Oregon, Illinois, Colorado, and points between. Introductions skipped the résumés. Nobody gave a talk. Each session opened with one question and a piece of pre-work everyone had already done. The rooms were small enough that everyone had to think out loud.
Everyone in the rooms was an adult, and the evidence was mostly memory — struggles that felt, at the time, like nothing but struggle, and turned out to matter. An adult sees a hard thing from the far side of it. A child is inside it, where it just feels hard. That limit runs through everything here.
We recorded everything. After each session we wrote a brief — not minutes, the sharpest ideas in the room — and sent it back for participants to argue with. Those four briefs are the source for this report. The quotes are real and unattributed. This is a group statement, and it belongs to the whole room. Once a draft report existed, the rooms met a fifth time to argue with that too. This version — leading the tool, not being led by it — is what came back.
Each session started with a question people in education are already asking. Each time, the room turned it into a better one. Those four turns are how we found the six capacities.
Four Questions In, Four Better Questions Out
Each chapter below is one turn: what we brought, where it flipped, what the room left behind.
1
The Foundation
The first room took the readiness question head on. What has to be there before a child reaches for AI? Not an age. “No AI before twelve” answers nothing. Not rules. Experiences.
The turn came early. AI augments what is already there. If nothing is there — no sense of self, no experience of struggle, no instinct for when an answer smells wrong — then AI is not augmenting. It is substituting. Readiness is not an age. It is a foundation, and you can name what goes into it.
“Where is the I in any of that?”
The room drafted six experiences a child should have before the tool.
- Productive struggle
- Being wrong, being stuck, and pushing through anyway. AI’s first move is to remove the struggle. If the struggle was the point, something has to build that capacity before the tool arrives.
- A sense of self
- Enough of an I to notice when something is not coming from it. This is a lifelong project, not a box checked at twelve. But some of it has to be forming before a child hands their thinking to a machine that writes in their voice.
- Critical questioning
- The instinct to ask whether something is true, and how you would know. A child with that instinct can survive an AI that makes things up. A child without it cannot.
- Human connection
- Real friendship, real conflict, real repair. A child who knows what real connection feels like has a reference point. Without one, the machine’s endless patience looks like the real thing.
- Answering yourself
- Finding something out on your own. Not knowing, held long enough, is where new thinking comes from. That muscle weakens fast once the answer is one click away.
- Teaching it back
- Explaining a concept to someone who does not understand it yet. A grade says you got the answer. Teaching says you know it.
The room also sketched what these look like in practice. Have a child catch the AI being wrong about something they know well. Have them prompt it to be unfriendly and notice that it feels nothing. Have them role-play telling it a secret, then follow where the data goes. Not a curriculum. The start of one.
2
The Mirror Problem
AI hands back a version of the person using it. It writes in their voice, finishes their thoughts, produces work that looks like theirs. For an adult with a settled sense of self, that is convenience. For a child still building one, it is something else. The second room went looking for what strong enough means.
Nobody could define a sense of self in the abstract. So the room flipped the question. What would you refuse to give away? Everyone had an answer. Therapy, because the human part is the medicine. Art, because expression has to come from somewhere. Deciding, because once the machine chooses, the choice stops feeling like yours. Hugging your mother. Work you care about.
The “never” list was sharper than any definition. Under every refusal was the same thing: a connection. To another person, to your own history, to your own taste. A child without those connections does not have a no to give. Forming one may be the precondition for everything else.
“What is it about AI that’s putting us in touch with our mortality in this way?”
The session produced three diagnostics instead of a test. Three questions you can put to a child, or to yourself, and get a real answer.
- What do you refuse to give away?
- Refusal is more articulate than definition. A child who can name something AI will never touch, and say why, has the beginnings of a self to protect.
- What are you losing without noticing?
- The slow forms of thinking go one shortcut at a time. The hunt for an answer. The mulling that lets a thought cohere over days. Not yet knowing. A child would not notice losing these. They would simply never develop them.
- Can you say why?
- Readiness shows in the defense, not the conclusion. Until a child can stand behind a choice — including the choice not to use AI — they are not using the tool. The tool is using them.
The room also threw out its own metaphor. An “AI driver’s license” implies a machine you decide to operate. You can fall into AI mid-search without deciding anything. What children need looks less like driver’s ed and more like self-defense. And self-defense, done right, is not a reactive crouch. One participant remembered a childhood kid-power class that taught her she could push back on an adult who spoke to her in a way she did not like — and anyone who trains this way will tell you the fight is the smallest part of it. It changes how you stand, what you notice, how you carry yourself all day. Not avoidance. Capacity.
3
The Struggle Question
Everyone got the same seventh-grade assignment before the session: a five-paragraph essay on whether phones should be allowed in class. One question. Can this assignment survive AI?
The room spent half an hour trying to save it. Debates judged live. Outlines written in front of the teacher. Closed-book defenses. Every fix worked. Every fix was expensive. And every fix conceded the same point. As written, sent home, the essay cannot be protected.
Then the frame turned over. The room had been protecting the wrong struggle. The teacher’s task is no longer “go home and write an essay.” It is “struggle with the things we actually want you to struggle with.” Say it that way and the essay stops being sacred. Maybe sentence structure still matters. Maybe building an argument matters more. Maybe the polished paragraph is exactly the part the machine should take. The question is not how to make children struggle the way they used to. It is which struggles are worth keeping.
“Choose your hard.”
The room set four conditions. A struggle is worth defending from AI when —
- It is chosen
- A struggle the child did not pick is one they will outsource at the first opportunity. As AI makes getting by easier, choosing which hard thing is yours starts to look less like an assignment and more like a coming-of-age. Which struggles are you choosing may be a better readiness question than can you struggle.
- It is toward something they value
- Nobody wants to struggle for its own sake, and they should not have to. A child will fail and retry endlessly at soccer, where the value is obvious, and refuse the same effort on an essay bound for a shelf. Give the work real stakes and the struggle gets taken up.
- It builds a muscle, not a product
- The artifact is disposable. The capacity is the point. This is what AI actually costs a developing mind. Not this essay. The reps that would have built the muscle the next hard thing needs.
- It leaves something theirs
- You can tell the struggle mattered if the child can point to the part that is their own. And stand behind it without the machine. If they cannot find themselves in the finished work, the struggle did not happen.
This room did not agree about everything. Asked whether struggle has value in itself or only as a means to something else, it split cleanly. The split is in Part Three.
4
The Social Mind
Learning is social. AI is private, patient, and agreeable. Before the session the fourth room went out to watch friction — a disagreement, a collaboration, two people stuck on a problem — and note what it produced. The reports matched. Two cousins fighting over rooms in a puppet show were surfacing a misunderstanding neither knew they had. Siblings arguing over a toy, with a parent mediating, got a rep in empathy. Friction is not the failure mode of social learning. It is the mechanism.
Then the frame flipped. Instead of asking how to protect children from a frictionless machine, the room asked what the machine does to the value of everything around it. Groups organize around whoever knows the most. If the expert is already in the room, and it is a machine, that reverses. The baseline answer is free. What a child is measured by is what unique insight they can add to the solution. One breakout said it in five words: AI is the new zero.
“What are the social skills required to withstand that ambiguity — in whatever time it takes between the world changing and the world changed?”
The room then catalogued what a patient oracle does not do. These are the things a child gets from people, and so far only from people.
- Friction surfaces assumptions
- You cannot find out you are wrong about the frame from something that keeps agreeing with you. The machine will happily assign the mediator role to a student, completely unaware that they are the one who started the argument.
- Teaching builds an identity
- When a child teaches, something happens to the child that has nothing to do with the content. The ten-year-old asked to teach his card trick did not just explain it. He performed it, put on a teaching voice, became the arbiter of the knowledge. Being the one who knows is a reward the oracle cannot give. It is always the teacher, never the student.
- The social burden is the training
- The thousand small awkward reps are the curriculum. Managing another mind’s expectations. Reading the person who has had enough. Working up the nerve to ask if you can sit next to someone. The machine’s frictionlessness is a relief. It is also the removal of the reps.
- A human teacher can be defied
- Some of the deepest learning is learning about the person teaching you. The minister who taught world religions without dogma reordered a student’s mind because of who was saying it. The machine shows up wearing nothing — no history, no expectation, nothing to overturn. For narrow content transfer it may win. The rest rides on the person.
The last round went somewhere the project had not gone out loud. Money. The value of expertise — the status it confers, the careers built on it — is a large part of what makes this technology frightening to adults, and pretending otherwise misses the anxiety under everything else.
The last round also produced the question worth keeping: If knowledge now arrives on demand, like water, what is the next way people are valuable to each other, and how do we teach it? Ready for the tool has to mean ready for the world the tool is making.
How the Four Became Six
Each room left its own sketch: six experiences, three diagnostics, four conditions, four things a patient oracle cannot do. Lay them on top of each other and they line up.
Productive struggle from the first room and the chosen struggle from the third became sit with hard things. The sense of self and the “never” list became know your own voice. Critical questioning survived intact as ask what is true. The first room’s human connection supplied the collaborate, and the fourth room’s friction supplied the collide — together, collide and collaborate with other minds. Students teaching as the proof of mastery came up in the first room and again in the fourth, from two different directions, so it became teach what you know. And AI is the new zero became add past the machine.
Six capacities. That is the card in Part One — and the fifth session gave the card its question. Every capacity on it, the room noticed, is answering the same thing from a different side: whether the person is leading the tool, or the tool is leading the person.
The Language the Rooms Made
Some of what a room produced is ideas. Some of it is language. These phrases did work in the moment and kept working after.
- Leading, or being led
- The fifth room’s compression of all six. Every capacity on the card is one way of answering the same question: who is deciding?
- Choose your hard
- Everything is hard, including outsourcing all your thinking. The skill is not avoiding difficulty. It is selecting it.
- The struggle muscle
- What hard problems build. The essay is disposable. The reps are not.
- Something’s gone
- A participant reading polished, AI-finished prose of his own and feeling the absence of himself in it.
- AI self-defense
- Better than “AI driver’s license.” A license implies a machine you choose to operate. Self-defense prepares you for something that may approach you before you are ready — and it is a practice, not a flinch.
- A patient oracle that never pushes back
- The tool, seen from a child’s side of the table.
- AI is the new zero
- The machine’s answer is the test already passed. From there, you are what you add.
- Bumpers on the bowling alley
- AI that gently facilitates children’s collaboration: safe, smooth, and hemming a child into the narrowest version of being human.
- Between the world changing and the world changed
- The interval we are preparing children for. Not the destination. The path they are going to be walking.
What We Are Still Arguing About
The rooms did not agree on everything. These are the live disagreements. We are carrying them forward because they are where the next thinking is.
Is struggle valuable in itself?
One side: struggle is a means, and once you have what it was for, it is just cost. Students offloading work to AI are not failing to value struggle. They are triaging a finite supply of effort toward what they care about most. The other side: struggle is growth, close to a tautology, and its absence shows up in children who meet difficulty and fold. The middle: we may be conflating struggling, suffering, and growing. Until we can tell them apart, “the value of struggle” is too blunt an instrument.
The machine cannot walk away.
Walking away is a communication. A partner who cannot do it is either infinitely safe or quietly mis-training. The room split. We are leaving it split.
Who this reaches.
The rooms so far skew fluent — households and schools already thinking about this. AI literacy looks different elsewhere, and a framework that only works for families who read reports like this one has failed. The same flag applies to children who show what they know differently. Any version of “defend it to a panel” has to make room for many ways of demonstrating a self.
Who gets to read the gauge.
An early reader heard, in this framework, a school deciding when a child has enough of a self — and named the cost: a child set back not by their own readiness but by an adult’s slow reading of them. That is the failure mode, and the children who show what they know differently would pay it first. We have tried to write the gate out of this document; we are not sure a document is enough. The honest version of the adult’s role is the one a parent already plays with the car keys — a long accumulated read, held humbly — never a certification. We are keeping this worry where we can see it.
How would we know?
The oldest question in education is still open here. Watching a child, what would tell you a voice is forming, or that a struggle is building muscle instead of just hurting? The rooms produced diagnostics, not measures — and any answer that arrives looking like a standardized instrument has probably already broken the thing it was sent to measure.
The teacher problem.
The pressure on children is also on the adults in front of them. Teachers are being pushed to use AI in ways that conflict with their own craft, and they are modeling that conflict live. Whatever gets built for children will land in classrooms where the adults are mid-mirror-problem themselves.
One size will not fit all.
A five-year-old on the spectrum has memorized his county’s entire transit system. The oracle’s patience with bottomless questioning is a gift no tired adult can match. The same child needs a grown-up nearby repeating: this is not a real person. The tool is both the best patient teacher he has and a relationship he needs help categorizing. Both are true.
Ask the children.
The people this framework is for have not been in the room. They are already struggling — with the moment, with the data, with the world they are inheriting — and nobody has asked them when struggle paid off and when it only hurt.
Where This Goes Next
The rooms have met five times now. The next step is not a sixth room of the same people. It is four doors out.
The people this is for
Nobody under eighteen has been in these rooms, and the framework knows it. We intend to put the card in front of young people and ask what works and what doesn’t. How could this help you and what needs to change to make it more effective. We will start with the bridge generation, the students who did school both before and after the tool arrived, who can sometimes name exactly what was lost and what was gained. Empathy interviews, youth councils, the networks already built to convene teenagers. If the six survive that contact, they will have earned it.
The people who would use it
Principals, teachers, program designers — with a harder question than does this resonate: what would you do with this on Monday? The places it could land are concrete: how systems get designed, how instruction gets built, how principals and teachers get trained, how a school writes the plan for one particular kid.
The science
What happened in these rooms was careful, structured, and honest — and it was still adults reasoning from memory about child development. Two things follow. First, map the six against what a century of developmental research already knows, and find out where we are rediscovering the wheel and where we are wrong. Second, push for the studies that do not exist yet: what sustained AI use actually does to the formation, and the retention, of exactly these capacities. That data is beginning to arrive. Somebody should be asking it these questions on children’s behalf.
The pressure
Waiting is cheapest for the protected. Plenty of adults can skip the tool for a week and lose nothing; plenty of young people are being told, more or less directly, use it now or be left behind — and that message does not land evenly across zip codes. A version of readiness that only works for families with room to wait has failed. Whatever this becomes next has to work under the pressure, not beside it.
If something here keeps coming back to you — a challenge, a story that makes one of these ideas concrete, a place you think we got it wrong — write to us at info@mirafellowship.org. A short email, a voice memo, a paragraph in passing. All useful.
And if there is a question in your world moving faster than your institutions can think about it, this format is built for that. A few small rooms, the right people, hard questions, and the discipline to publish what surfaced. That is a Mira Rethink. There will be more.