My 10-Year-Old Vibe Codes. She Also Does Math by Hand. Why That's the Only Strategy That Works.

According to one of the most popular scientific magazines in the world, artificial general intelligence is here now today. But none of us have figured

now today. But none of us have figured out what we're going to teach our kids.

This is a real issue. We can use a tool like Claude Code to build an entire medical school curriculum in just two weeks. And I'm not saying that because

weeks. And I'm not saying that because it's a hypothetical. Someone shared that directly on X. 450 lectures, 16,000 figures, roughly a 100 million tokens of automated work with multiple rounds of

error checking, and 99% of it was flawless. Work that normally takes

flawless. Work that normally takes hundreds of faculty years to produce was done in just a couple of weeks by one person. Meanwhile, two billion kids

person. Meanwhile, two billion kids around the world are enrolled in schools that have no idea any of this is possible, that are running on an educational philosophy designed at best

for a 20th century industrial economy.

that economy is not going to exist when these kids get to adulthood. And this is personal to me. I have kids, too.

Nature, the journal, not the concept, published a peer-reviewed argument that said that artificial general intelligence has indeed arrived. And I

quote, "The machines Touring envisioned 75 years ago have arrived." Globally,

86% of students report using AI in their learning, according to the digital education council. In the UK, the

education council. In the UK, the picture is even more dramatic. usage

surged from 66% in 2024 to 92% just the next year in 2025 per HPI's annual student survey. The point is growth,

student survey. The point is growth, right? AI tutors outperform human tutors

right? AI tutors outperform human tutors in controlled studies. And nobody, not schools, not governments, not parents, has figured out what to teach our kids and what comes next in a world that's

changing this fast. I have three kids. I

work in AI every day. I've spent weeks figuring out how to talk about the latest in AI. From autonomous agents that negotiate car purchases to dark factories where code writes itself to

enterprise adoption curves that frankly looks like hockey sticks drawn by a drunk person who just wants to draw the line straight up. That's the world we're living in. And yet in the evening I go

living in. And yet in the evening I go home to the same question every parent is facing whether we realize it yet or not. What does education mean when

not. What does education mean when machines can do most of what we spent the last century teaching humans to do?

My 10-year-old sat at the kitchen table last month working through long division by hand with a pencil because I asked her to do that. I'm also teaching her to vibe code with Claude. These are not

contradictory positions. They're the

contradictory positions. They're the only positions that make sense together and the reasoning behind them is what I want to talk about in this video. Before

we get into that, let's talk about the world that our kids are going to inherit. A Harvard study published last

inherit. A Harvard study published last year found that students using AI tutors learned more than twice as much material in less time than students in traditional settings. A collaboration

traditional settings. A collaboration between Ed and Google DeepMind showed AI tutoring systems outperforming human tutors on problem-solving tasks, 66% versus about 60%. When you combine human

teachers with AI tutoring, the knowledge transfer doubles. So Benjamin Bloom

transfer doubles. So Benjamin Bloom established decades ago that one-on-one tutoring produces a significant improvement about two sigas in standard deviation. That's a massive effect. The

deviation. That's a massive effect. The

constraint was never whether personalized tutoring works. We figured

that out decades ago. The constraint was always that you can't give every child a personal tutor. AI is removing that

personal tutor. AI is removing that constraint today. And that is part of

constraint today. And that is part of what our kids are inheriting. Khan

Academyy's AI tutor, Kmigo, went from 68,000 users to 1.4 4 million in just a year. 266 school districts in the US are

year. 266 school districts in the US are being served by Khan Academy today. Saul

Khan calls it probably the biggest positive transformation that education has ever seen. An 8-year-old can build video games with Claude today by typing instructions like make the bad guys

tigers. Make them move slower. A mom

tigers. Make them move slower. A mom

with no coding background can build a personalized AI tutor for her dyslexic son using vibe coding. Just natural

language and iteration, nothing else.

Zack Yadagari, 18 years old, is the CEO of Cal AI, and he's pulling down 1.4 million a month right now with 8.3 million app downloads because he decided

to be entrepreneurial and use AI to build an app that serves customers. A

13-year-old from Toronto met Sam Alman at a tech conference because he'd built something worth showing at 13. This kind

of environment is the water every kid on Earth is swimming in. Pretending it's

not there doesn't make them better swimmers. And handing them a jet ski

swimmers. And handing them a jet ski before they've learned to swim doesn't make sense either. This is like the calculator moment, except it's for everything. Back in the 1970s, when

everything. Back in the 1970s, when electronic calculators became more affordable, the education establishment panicked. Calculators in classrooms were

panicked. Calculators in classrooms were considered cheating. Full stop. They

considered cheating. Full stop. They

would destroy children's ability to do arithmetic. They would produce a

arithmetic. They would produce a generation that could not think mathematically. So, schools banned them.

mathematically. So, schools banned them.

Parents protested them. And while we may not remember it now, the debate consumed education policy for over a decade. We

know how that ended. Calculators did not destroy mathematical thinking. They

changed what mathematical thinking meant. Once students didn't need to

meant. Once students didn't need to spend 20 minutes on long division, they could spend that time on the concepts long division was supposed to serve.

Proportional reasoning, algebraic thinking, problem decomposition. So the

tool ended up freeing the learner from the mechanical to engage with the meaningful. And schools typically have

meaningful. And schools typically have decided they want to have a foundation in mechanics and then move kids into calculators over time. Here's the part of the calculator story that gets left out. The transition worked because

out. The transition worked because students did still learn those mechanics first. They understood what the

first. They understood what the calculator was doing. They could

estimate whether an answer was reasonable or not. They could catch errors. They had the foundation and the

errors. They had the foundation and the tool extended it. So the parents who said calculators will make our kids stupid, and there were lots of parents in the 70s who thought that, they were wrong. The parents who said just give

wrong. The parents who said just give them calculators and skip the math, they would have been wrong, too, if anyone had been extreme enough to suggest that the right answer turned out to be both.

Build the foundation and then give them the tool. We're in that calculator

the tool. We're in that calculator moment again, except it's not just arithmetic. It's reading, it's writing,

arithmetic. It's reading, it's writing, it's research, it's analysis, it's coding, it's creative work, it's communication, it's problem solving.

Every single cognitive task that AI can now perform competently. The scope is fundamentally different from 1975, but I think the principle is not. So, let's go

back to basics. Why am I insisting on my kid doing long division by hand in a world of AI? My kids read real books, physical books, not screens. We do math by hand before we do math with tools. We

write with pencils on paper. The

reasoning connects directly to everything I've been writing about all year. The single most important finding

year. The single most important finding from watching autonomous agents operate in the real world is that the quality of the output is determined by the quality of human specification. I've talked

about that a lot just in the last couple weeks with the Open Claw moment with Opus 4.6. Human specification matters

Opus 4.6. Human specification matters more and more. I wrote about a Maltbot agent that negotiated $4,200 off a car purchase while its owner was in a meeting. That same week, a different

meeting. That same week, a different agent sent 500 unsolicited messages to friends and family and the developer's wife. Same technology, same

wife. Same technology, same architecture. And what I called out is

architecture. And what I called out is the difference is the human's ability to specify. If you have clear objectives,

specify. If you have clear objectives, if you have defined constraints, if you have a bounded communication channel, then you are in business. If you have broad access and vague boundaries, if you can't specify, you're in trouble.

That's a human skill and it's practiced manually and it's something we can teach our kids. That insight scales to

our kids. That insight scales to education really, really easily. You

don't get to write a good spec for something you don't understand. You

can't evaluate an AI's output in a domain where you have no knowledge. You

cannot exercise good judgment, things like taste, like discernment, like critical thinking about work that you've never engaged with deeply enough to internalize. So, when my kid asks Claude

internalize. So, when my kid asks Claude to help with a math problem, I want her to know enough to recognize when Claude is wrong. Last month, Claude confidently

is wrong. Last month, Claude confidently worked her through a word problem and arrived at an answer that did not pass a sanity check for me. When my kid is older and she can use Claude for math, I

want my kid to know enough to recognize when Claude is wrong. When she uses Claude for coding, I want her to know enough to recognize good separation of concerns as an architectural principle.

These are human skills first and we leverage them with AI tools later. The

Harvard tutoring study, the one that showed AI tutors can double learning outcomes, found that the best results come from human AI collaboration, not from replacing the human with AI. The

human needs to bring something to that collaboration and that something is the foundation I'm talking about here.

Reading physical books builds mental models that no AI can build for you passively. Not because AI can't explain

passively. Not because AI can't explain what Moby Dick means, but because the cognitive work of reading, of struggling with the text, of rereading, of integrating the ideas is itself the

learning a human brain needs. The

struggle is the point. Math by hand builds a sense of numbers you don't get any other way. An intuitive feel for magnitude, for proportion, for relationships that shortcuts any bypass

you can get from talking with Chat GPT about statistical distributions. Writing

by hand builds the connection between thinking and expression that typing and dictation tend to compress in ways that affect our ability to remember and our ability to comprehend. None of what I'm saying here means AI is bad for

learning. Quite the opposite. The

learning. Quite the opposite. The

evidence suggests it's great. The

evidence shows that if AI can help extend onetoone tutoring principles, we should do it. But it also means the foundation comes first and the foundation is built through effort with

our human brains through learning discipline as kids not through efficiency. Look, I am not in the

efficiency. Look, I am not in the protect the children from AI camp. I

have watched my kids vibe code websites and I love it. What I see when kids use AI is not intellectual laziness. It's a

different kind of intellectual work and a genuinely valuable one that we should encourage. Last week my kid wanted

encourage. Last week my kid wanted enemies in a game that she's building.

So, she typed add enemies. Claude added

enemies. Enemies that can spawn off screen, move in the wrong direction, can't be hit. It doesn't work, she said.

So, we talked about it and I asked her what she really wanted the enemies in the video game to do. And she thought about it and then she said, "Add three enemies that spawn from the right side of the screen, move them left at about a

medium speed, and make them disappear when the player touches them." Suddenly,

she got the behavior she was looking for. And that little conversation taught

for. And that little conversation taught her more about spec quality than a lesson that I could have scripted. When

a kid vibe codes, they're doing several things at once. They're specifying

requirements in natural language for something they are interested in doing.

They're decomposing a complicated vague desire into discrete tasks and they're learning to iterate, test the result, see what doesn't match, refine the specification. They're not debugging

specification. They're not debugging code really. They're debugging their own

code really. They're debugging their own intent. And these are skills that

intent. And these are skills that transfer. They map directly onto what

transfer. They map directly onto what professional software development and building products for customers increasingly looks like regardless of what job title you have. And they

develop a capacity for precise thinking that transfers well beyond coding. Andre

Carpathy, Tesla's former head of AI, one of the architects of the deep learning revolution. He founded Eureka Labs

revolution. He founded Eureka Labs specifically to build what he calls an AI native school. His stated goal is to raise young people who are proficient in the use of AI but can also exist without

it. That formulation is one that I keep

it. That formulation is one that I keep coming back to. You need to be proficient and also independent, not one or the other. Carpathy also said something that I think every parent and

teacher needs to hear right now. Quote,

"You will never be able to detect the use of AI in homework. Full stop. He's

right. The arms race between AI writing detection and AI writing generation was over before it started. And the people who are selling AI detection to schools

are wrecking the outcome of kids. They

are judging kids based on a huristic that is mathematically impossible to implement in a correct way. You cannot

detect AI in homework. Full stop. The

educational response cannot be better detection. And I see too many cases

detection. And I see too many cases where students are being pushed out of school for something they did not do because of a tool that administrators blindly believe is accurate just because

of the messaging on the tin. You can't

detect AI writing. Full stop. It has to be a fundamental rethinking of what we're measuring and why, which is long overdue anyway. If a mother with no

overdue anyway. If a mother with no programming can build a personalized AI tutor for her son, which is true, real example, if you can tailor the reading experience to that child's specific

needs, as a mom with no coding experience, you're not replacing education in that sense. You're

extending education to a child the traditional system is failing. The tools

can reach children that our current institutions cannot or won't or don't know how to do. And that is a much more worthwhile use of AI for our education system than trying to purchase from

vendors selling snake oil, promising they can detect AI homework. The

families pretending AI don't exist are making the same mistake as the schools that banned calculators in 1975.

Technology will not go away. The

children who don't learn to use it critically, skillfully, as a tool under their direction rather than a crutch will fall behind in ways that compound every single year. The skill connecting

foundation to AI fluency is metacognition. The ability to think

metacognition. The ability to think about your own thinking, to know what you know, to know what you don't, and to make deliberate decisions about when to rely on yourself versus when to delegate

to a tool. Researchers increasingly call this the defining competence of the AI age. Not what you know, not what the

age. Not what you know, not what the machine knows, but your capacity to move between the two. strategically

allocating your cognitive efforts, coordinating AI assisted tasks, and evaluating results against your own understanding. In practice, it's the

understanding. In practice, it's the difference between a kid who asks chat GPT to write the essay and a kid who drafts the essay, uses AI to identify the weak argument, strengthens them with her own thinking, and produces something

neither she nor the AI would have created alone. The first kid, they

created alone. The first kid, they completed an assignment. The second kid learned something and created something genuinely new. Same tool, different

genuinely new. Same tool, different metacognition skill. This maps directly

metacognition skill. This maps directly to the themes I've been writing about all month. Agency, the ability to direct

all month. Agency, the ability to direct AI rather than be directed by it. Taste,

knowing what good looks like when the machine can produce infinite mediocrity at zero cost. Specification quality,

articulating what you want precisely enough that a system can execute it well. These aren't actually technical

well. These aren't actually technical skills. They're cognitive skills with

skills. They're cognitive skills with technical application. and they develop

technical application. and they develop the same way every other cognitive skill develops through practice, struggle, feedback, and gradually increasing the challenge. Singapore's AI education

challenge. Singapore's AI education framework captures this as a progression. Learn about AI, learn to

progression. Learn about AI, learn to use AI, learn with AI, learn beyond AI.

That last step, learn beyond AI, matters most, and I don't think anybody has figured out how to teach it systematically yet. is where the student

systematically yet. is where the student doesn't just use the tool, but transcends the tools limitations through their own judgment and creativity. I

don't think that step gets solved in a classroom very well. I think it gets solved at kitchen tables in conversations with our kids about what AI got right and what it got wrong and why. in the moments where we ask them to

why. in the moments where we ask them to try it themselves before asking the machine. There's a concept in psychology

machine. There's a concept in psychology called learned helplessness, where a person repeatedly experiences situations where their own effort doesn't matter, where outcomes are determined by forces outside their control, and they

eventually stop trying. Not because

they're lazy, but because their brain has learned the effort doesn't matter.

The AI version of this plays out through what researchers call cognitive offloading. You delegate a mental task

offloading. You delegate a mental task to a tool. The tool takes care of it.

Over time, the neural pathways that would have handled the task don't really develop or if they existed, they weaken.

The offloading becomes a kind of dependence. The dependence becomes

dependence. The dependence becomes helplessness. And that's not necessarily

helplessness. And that's not necessarily a dramatic moment. It's happens

gradually. It's not sudden. It's a quiet erosion of capability that comes from never needing to exercise that skill.

This is not theoretical. Educators are

reporting it in real time. College

professors are describing today that students are arriving in the classroom who can no longer read a full chapter, who can no longer synthesize an argument from multiple sources or sit with a

difficult text long enough to extract meaning from it. High school teachers report that writing quality has absolutely collapsed. Not just because

absolutely collapsed. Not just because students submit AI generated work, although many do, but because even the students who aren't using AI have lost the habit of struggling through a draft,

the muscle has atrophied before we really noticed it was weakening. A

growing number of faculty are redesigning their courses around in-class work and oral exams because take-home assignments have become functionally meaningless as a measure of capability. The phrase I keep hearing

capability. The phrase I keep hearing from educators is they can't do it anymore. Not won't, can't. That is the

anymore. Not won't, can't. That is the evidence I'm responding to when I sit my kids down with pencils and paper. It's

not about nostalgia. It's a direct response to what's happening to the first generation of students who had AI available and who never built the foundation to function without it. I'm

not willing to let that happen at my kitchen table. And this foundational gap

kitchen table. And this foundational gap extends into the emotional domain. 3/4

of teenagers are now using AI companion chat bots for emotional support. Not as

a supplement to human relationships, but in some cases as a primary source of emotional connection. The chatbot is of

emotional connection. The chatbot is of course always available. It's always

patient. It never judges. It never makes demands. It also isn't real. It cannot

demands. It also isn't real. It cannot

teach conflict resolution because there's no genuine conflict. It can't

build relational resilience because it never pushes back when the stakes are real. It can't model empathy because it

real. It can't model empathy because it has no experience to draw upon. Multiple

tragedies have been linked to these kinds of parasocial relationships that teenagers are having more and more across the country and the world. That's

the extreme. The everyday version is subtler and it's much more pervasive.

The AI is so helpful, so frictionless, so immediately gratifying that reaching for it becomes the default before a child has to try to think through the problem on their own. That's true for

math as much as emotions. Every time

they choose that path, the harder path gets a little bit more difficult. Not

because AI is actively making our kids dumber, but because it's making it easy to not take the difficult route where the actual learning lives. And so it

stops feeling worth the effort. So the

trap is not really that AI will be too powerful or that it will take over. The

trap is that it will be so seamlessly and perpetually helpful that your kids and mine are never going to develop the tolerance for difficulty that real learning requires. They're going to end

learning requires. They're going to end up producing impressive looking work without understanding it deeply enough to defend it, to extend it, to know when it's wrong. I am convinced the answer is

it's wrong. I am convinced the answer is not going to be withholding these tools.

The answer is going to be sequencing AI tooling directly and deliberately into the education system. foundation first,

learn with your head first, struggle first, and build the muscle before you add that AI exoskeleton that extends your capabilities. And once that AI

your capabilities. And once that AI exoskeleton is on, you have to keep exercising without it so the muscles don't atrophy. And that's true for us

don't atrophy. And that's true for us adults, too. So, what am I actually

adults, too. So, what am I actually doing with my kids? I don't have a curriculum for this. Neither does

anybody else really. The world is moving faster than any educational framework can really track. Singapore is trying.

They're rolling out AI training for teachers at all levels next year.

Finland has national recommendations.

44% of homeschool parents are already using Chat GPT in their teaching.

Frankly, a higher rate than classroom educators. Everybody's improvising.

educators. Everybody's improvising.

Here's how I approach it. As I've been sharing, the basics are non-negotiable.

You got to read real books, not summaries, not audiobooks at 2x speed.

Real reading for kids with real cognitive effort. math by hand until the

cognitive effort. math by hand until the concepts get internalized and not just performed until you understand numerousy. Writing with a pencil because

numerousy. Writing with a pencil because the physical act of forming letters engages the brain differently than typing and that difference matters for memory and comprehension. These are not

romantic preferences for an oldtimey wood cabin lifestyle. They're

investments in cognitive infrastructure that makes everything else possible down the road. And then yeah, I introduce AI

the road. And then yeah, I introduce AI tools. I introduce them deliberately,

tools. I introduce them deliberately, not as a default, but as an extension. I

talked about vibe coding together. We

code games. We solve problems. We iterate on designs. I ask my kids to explain what they want before they ask the AI. I make them articulate the goal,

the AI. I make them articulate the goal, the constraints, what good will look like, and then I ask them to critically evaluate what comes back. I ask them to be directors, not audience. My

10-year-old is going to need to get into agents. I know this because I've spent

agents. I know this because I've spent weeks writing about what agents can do and the capability curve is accelerating, not flattening. Her world

is going to be agents. But agents are the most spec dependent technology we humans have ever built. The entire value proposition depends on the human's

ability to define a task precisely, set appropriate constraints, and evaluate the output critically. Teaching a kid to use agents before she can think clearly about what she wants would be like handing her the keys to the car before

she can read the map. So I think about it in terms of a readiness model. My

10-year-old is somewhere around steps two and three. She's building games with Claude. She's learning to articulate it.

Claude. She's learning to articulate it.

So the progression looks like this.

Build cognitive foundations. Introduce

AI tools with guidance. Practice

directing AI with increasingly clear spec. and then graduate to agent level

spec. and then graduate to agent level autonomy as judgment starts to develop.

That's not really a timeline. It's more

of a readiness model. And I'll be honest, my 10-year-old right now is between steps two and three. She's

building games with Claude and learning to articulate what she wants, but she's not at a point yet where she would be ready for agent level autonomy. These

are real and valuable skills. We're not

going to rush through them, and we're in the middle of the progression, not the end. I watch for signs of cognitive

end. I watch for signs of cognitive offloading. When one of my kids asks

offloading. When one of my kids asks what AI would say before trying to think through a problem, I redirect not with a lecture, but with a question. What do

you think the answer might be? Go

research in the encyclopedia. Not

because I don't want them using AI, but because I want the AI to extend their thinking, not replace it. Seymour

Pepert, the MIT researcher who pioneered computational thinking in education, wrote in 1968 that computer programming gives children a way to think about their own thinking. He called it

constructionism, the idea that people build knowledge most effectively when actively making things in the world. Not

consuming information, but constructing with it. His vision took 50 years to

with it. His vision took 50 years to reach the mainstream. And by then, tech had outrun his framework. And we're not really talking about the way he talked about computing at the time. But his

core insights hold now like they did then. Children learn by building. The

then. Children learn by building. The

act of creation through computing helps kids develop the cognitive architecture that matters. An 8-year-old building a

that matters. An 8-year-old building a game with Claude is taking constructionism at a scale that Paper never envisioned. The creation is real,

never envisioned. The creation is real, even if the code was written by a machine because the thinking is the child's. I've done some work to distill

child's. I've done some work to distill all of this down into seven principles that I think scale for parents who are struggling with education and I want to share them with you here because I want

us to find ways to help our kids to be human even in a world where technology is changing faster than ever. Principle

number one is foundation before leverage. I've talked about that a lot

leverage. I've talked about that a lot in this video. Reading, math, writing.

Lean into the effort here. Not because

AI can't do these things, but because your kid can't evaluate what AI produces without understanding the domain.

Principle two is specification is the new literacy. The gap between a good AI

new literacy. The gap between a good AI outcome and a catastrophe is the quality of the human specification. Teach kids

to say what they want, the goal, the constraints, what done looks like. If

you can write a clear spec for an AI, you are kind of exercising the same muscle as if you're learning to write an essay. Principle three, be a director,

essay. Principle three, be a director, not a passenger. When your kid uses AI, they should be defining the ask, the task, the output, what to keep, revise,

and reject. If they're just passively

and reject. If they're just passively consuming what the AI produces, they're not learning, they're outsourcing.

Principle number four, sequence the autonomy. Start with bounded educational

autonomy. Start with bounded educational tools that have guard rails and graduate eventually to open-ended tools with guidance. You can vibe code together.

guidance. You can vibe code together.

You can build projects side by side, then get into agent level autonomy later. This progression should follow a

later. This progression should follow a degree of cognitive readiness. It's not

really age gated, right? Some

12-year-olds might be ready for agents.

Sometimes adults are not. Principle

number five, teach kids to catch the machine. AI will be wrong. Confidently,

machine. AI will be wrong. Confidently,

fluently wrong. Train kids to sanity check outputs against their understanding. When the AI makes an

understanding. When the AI makes an error and the kid catches it, that's not a tool failure. The conversation needs to be about how we built a foundation so we can critique and understand and

evaluate machine outputs together.

Principle six, build don't browse.

Making things with AI develops cognition in ways that consuming AI output does not. Videoding a game, designing an app,

not. Videoding a game, designing an app, creating art, these are active choices.

Asking the AI to summarize a chapter, that's passive. But Parrot was right.

that's passive. But Parrot was right.

Construction is how kids learn.

Prioritize creation over consumption every single time. Principle seven, last but not least, attempt before augmenting. I think this is the most

augmenting. I think this is the most important habit you can build. Try it

yourself first and then use AI to extend what you've started. Ask, "What do you think the answer is?" and then go, "Well, what does Chad GPT think?" The

kid who drafts before she edits is going to be learning in a way that the kid who prompts before she thinks is not. I

don't think these are permanent answers.

I think all of them will probably need to evolve as the tech does, but they're the best and most stable operating principles I found for raising kids who can direct intelligence rather than

depend on it. And to be honest, these principles are useful for learning AI as an adult as well. Look, I don't know the right ratio between foundation and fluency. Nobody does because the AI is

fluency. Nobody does because the AI is evolving so fast. The foundation

obviously matters more at six than 16.

Tool fluency is going to matter more at 16 than six. The parents who ban AI outright and the parents who hand over the iPad and walk away are both choosing comfort. One is the comfort of familiar

comfort. One is the comfort of familiar education and the other is the comfort of not engaging with a problem that doesn't have a great answer. This AI

problem is not going to get easier. The

world is changing faster than any curriculum can track. And anyone who tells you they've figured out what to teach the kids is selling you something.

What I have are the principles and daily practices I've rooted in as a parent.

You got to build the brain first. You've

got to give that brain leverage with AI.

You've got to teach your kids to think and then teach them to direct. Give them

the struggle, the gift of struggle that develops real capability and challenge yourself in the same way. Watch closely

enough to know when the tools are helping and when they're replacing the work that needed to happen in their heads. That's where I'm at. I'm at the

heads. That's where I'm at. I'm at the kitchen table. I'm doing long division

kitchen table. I'm doing long division worksheets. I'm using my brain, too, on

worksheets. I'm using my brain, too, on that. By the way, Claude might be open

that. By the way, Claude might be open on the laptop, but Claude doesn't get asked first. I try to raise humans who

asked first. I try to raise humans who can thrive in a world where none of us fully understand it yet. And it's the hardest problem that I have to work on, to be honest. It doesn't resolve easily.

I have to keep going after it day by day. The machines that Turing Envisioned

day. The machines that Turing Envisioned have arrived. Nature magazine was right.

have arrived. Nature magazine was right.

And our kids have to be able to do the work in the moment to get their brains ready to partner with machines that are truly intelligent because anything else

would be a disservice to them and frankly a disservice to us as their parents, as the ones charged with helping them find their way in a world that is going to look nothing like the world where you and I were kids.

Everyone on Earth is living through this transition. And if you can realize it

transition. And if you can realize it and help your kids navigate it, you may just be the person who helps someone

navigate the AI transition with so much less angst than you yourself feel.

That's one of the gifts we can give our kids. I look at my kids and I know they

kids. I look at my kids and I know they are not going to feel the same angst of transitioning through this world as I do and you do. They never knew a world before this one. Our gift to them is to

take the best of what we know in terms of cognitive architecture and cognitive skills, culture, make sure that transitions to them. Make sure they learn that. Make sure they put the

learn that. Make sure they put the effort in. Frankly, it's a gift to

effort in. Frankly, it's a gift to ourselves. If you haven't picked up a

ourselves. If you haven't picked up a book lately, pick up a book lately. It's

good for your brain. Make sure you're putting in the time to keep your own brain healthy so that you're using AI effectively. One more note. If you are

effectively. One more note. If you are not a parent, this applies to you as an aunt or an uncle of kids in your lives.

And this applies to you as a learner.

Don't forget, we all have the responsibility to bring up the next generation in a way that can effectively partner with machine intelligence. It

really will take a village more than it has in any previous generation. Best of

luck out