We May Be Educating Children for a World That Just Ended

What happens to education when intelligence is no longer scarce?

I’ve had one sentence stuck in my head since I heard Sam Altman say it:

“A kid growing up today will never be smarter than AI.”

Just sit with that for a second. Not because Sam Altman is necessarily right about every prediction. And “smarter” is obviously a messy word. Intelligence isn’t one number. Human beings aren’t benchmarks. But forget the semantics for a minute. What if he’s directionally right?

What if the kids entering school today spend their entire adult lives surrounded by intelligence that knows more than they do, remembers more than they do, can reason across more information than they can, and is available essentially on demand? Because if that’s even close to where we’re going, I think we have a much bigger question than whether students should be allowed to use ChatGPT on their homework.

We may be educating them for the wrong world.

And before anyone reads that as an argument against teachers, schools, professors, or education itself, it isn’t. Actually, quite the opposite. Next to my parents, some of the greatest influences on my life have been teachers and professors. Hands down. There are people who taught me decades ago whose voices I can still hear when I’m working through a problem. They shaped what I became curious about, what I thought I was capable of, and in many cases how I learned to think in the first place.

So this isn’t about throwing away where we came from. It’s about being honest about where we’re going.

Education was built for a world where intelligence was scarce

For thousands of years, becoming capable meant putting intelligence inside yourself. You studied. You memorized. You practiced. You accumulated knowledge and experience until eventually you became the person in the room who knew what to do. We built an entire educational system around that reality. And it made perfect sense.

If you wanted to become a lawyer, engineer, physician, accountant, architect, programmer, scientist, or almost anything else requiring deep expertise, there was really no shortcut. The knowledge had to live inside you because you were the thing doing the thinking. You were the processor. You were the memory. You were the reasoning engine.

So education naturally optimized around putting more capability inside the individual. Learn the material. Memorize the vocabulary. Master the formula. Practice the procedure. Produce the correct answer. Become the expert. That model worked because expertise was scarce and expensive.

Then, almost overnight on the scale of human history, something weird happened. Intelligence started showing up outside of us.

Not information. We’ve had books forever and search engines for decades. I’m talking about something qualitatively different. Something that can take information and reason across it. Challenge it. Explain it. Write with it. Code with it. Translate it. Find patterns in it. Generate hypotheses from it. Critique its own output. And increasingly, go do something with it.

That’s a different thing. And I don’t think education has really confronted what it means yet.

Maybe the students aren’t the problem

Right now we’re having arguments about students “cheating with AI.” And sure, sometimes they’re cheating. If the assignment is meant to teach you how to form an argument and you dump the prompt into an AI and paste the answer, you probably didn’t learn much. That’s real.

But I keep wondering whether we’re focusing on the easiest part of the problem because the larger question is much more uncomfortable.

We give students questions for which answers already exist. We isolate them from tools they will later use professionally. We reward recall. We measure individual execution. Then we graduate them into a world where they will increasingly work alongside intelligence that can perform many of those tasks instantly. There’s a disconnect there.

And no, I don’t think the answer is a generation of kids who know nothing because “AI knows it for me.” That’s probably just as bad as pretending the technology doesn’t exist. They still need to know things. A lot of things.

You can’t call bullshit on an AI if you don’t know enough to recognize the bullshit. You can’t recognize a subtle error in an engineering recommendation if you’ve never learned engineering. You can’t challenge a historical claim if you have no understanding of history. You can’t ask sophisticated questions in a domain where you have zero mental models.

Foundational knowledge still matters. Deep expertise still matters. Maybe more than ever. But perhaps we’ve confused the raw material of an educated mind with the final product.

What is all that learning actually for?

This is the question I keep coming back to. Historically, education often worked something like this:

I know this, therefore I can do this.

The knowledge and the capability were tightly connected because the knowledge had to be stored inside the person doing the work. But the emerging model might look more like:

I understand this deeply enough to question it, reason about it, combine it with other ideas, direct intelligence toward it, and judge what comes back.

Those are very different outcomes. And this is where the conversation gets interesting, because AI may simultaneously make knowledge less scarce and deep knowledge more valuable. That sounds contradictory. I don’t think it is.

If almost anyone can produce an answer, producing an answer becomes less valuable. Knowing whether it’s a good answer becomes more valuable. Knowing what assumptions are buried inside it becomes more valuable. Knowing what the machine missed becomes more valuable. Knowing that the original question was wrong becomes much more valuable.

The more capable the machine becomes, the more consequential human judgment becomes.

Maybe we’re measuring human capability the wrong way

Imagine telling a kid that the goal of physical education is to eventually become stronger than a bulldozer. It’s ridiculous.

We accepted a long time ago that machines could possess vastly more physical power than human beings. A forklift can lift more than you. A car can run faster than you. A crane can move something you could never move. We didn’t respond by declaring human beings obsolete. We stopped defining human value by horsepower.

I wonder if AI eventually forces us to make the same adjustment around cognitive horsepower. If a machine can search more information, remember more facts, calculate faster, translate more languages, generate more possibilities, and outperform an individual human across certain intellectual tasks, maybe the goal isn’t to defeat it at those things. Maybe those capabilities become leverage.

The question changes from How smart are you? to What can you do with access to intelligence this powerful?

That sounds like a small wording change. It isn’t. It’s an entirely different educational philosophy.

Curiosity stops being a “soft skill”

I’ve been writing recently about what I call The Age of Imagineering, a period where increasingly abundant intelligence begins reducing the friction between imagining something and actually creating it. And I think education sits right in the middle of that transition. If execution gets dramatically cheaper, then the ability to decide what is worth executing becomes more valuable.

Curiosity matters. Judgment matters. Taste matters. Agency matters. Perspective matters. Being able to see a problem everyone else walks past matters. Being able to connect something you learned in biology with something you learned in economics with something you experienced personally ten years ago matters. Being able to look at the first answer and say, “No. That’s not good enough. We’re solving the wrong problem,” matters.

For most of human history, an enormous amount of imagination simply died inside people’s heads. You could have a fantastic idea and still lack the capital, technical expertise, people, infrastructure, or access necessary to pursue it. That barrier is weakening.

A student growing up today may eventually be able to describe a software product and have most of it built. Ask a scientific question and run thousands of simulations. Imagine a physical product and coordinate systems that design, source, manufacture, test, and ship it. Direct the equivalent of teams of analysts, developers, researchers, and designers without personally possessing every skill represented inside those teams.

That changes what agency means. It changes what creativity means. And I think it changes what curiosity means. Curiosity starts looking a lot less like a personality trait and a lot more like productive capacity.

But there is a trap here

This is where AI enthusiasm can go too far. There is a very tempting argument that goes something like this: “If AI can do it, why should the student learn it?”

Because sometimes the struggle is the education.

A child working through a difficult math problem isn’t just trying to arrive at the number at the bottom of the page. They’re constructing something inside their brain. A student wrestling with an essay isn’t merely producing 1,500 words. They’re learning how to organize thought. Someone debugging code manually is often developing intuition that becomes extremely valuable later.

If AI removes every difficult step before those mental structures are built, we’ve created a strange kind of capability. The student can produce the artifact. But maybe they can’t actually do the thinking underneath it. That’s dangerous.

So I don’t think the future is simply “let the AI do everything.” The interesting challenge is figuring out: Which friction develops the mind, and which friction is just friction? Those are not the same thing.

We eventually stopped requiring professional engineers to perform every calculation manually. Scientists don’t physically wander library stacks every time they need a journal article. Software engineers don’t rewrite every primitive from scratch. Nobody thinks an architect is less of an architect because they use modern design software. At some point, tools become part of the profession. Education has to figure out when that transition should happen.

Maybe the model is something like aviation. We don’t teach pilots that automation is evil. We also don’t teach them to blindly trust automation. They learn how flight works. They learn what the instruments are telling them. They learn what failure looks like. They build enough understanding that when the system starts behaving strangely, they can recognize it. Then they use the technology.

Teach the mind first. Then teach the mind how to fly with machines.

What if the test is testing the wrong thing?

Here’s one thought experiment I can’t shake. Imagine the most consequential exam of a student’s life. Now give that student access to the most powerful AI available. Not to make the exam easier. Make the problem much harder.

Give them something messy. No obvious correct answer. Conflicting evidence. Incomplete information. Competing incentives. Give them an AI that occasionally produces brilliant analysis and occasionally produces extremely persuasive nonsense. Then watch what happens.

• Can the student figure out what the actual problem is?
• Can they ask better questions as they learn more?
• Can they challenge the model?
• Can they recognize when the evidence doesn’t support the answer?
• Can they find primary sources?
• Can they deliberately seek an opposing argument?
• Can they connect ideas across disciplines?
• Can they recognize that the original premise was flawed?
• Can they explain why they accepted one recommendation and rejected another?
• Can they ultimately produce something useful?

Now compare that with asking the same student to sit alone in a room and reproduce information from memory without access to any of the tools that will surround them during their professional life. Which one tells us more about their future capability?

I genuinely don’t know exactly where that line should be. There are absolutely things people should know without assistance. We need to know whether someone has actually built a functioning mind before we celebrate what that mind can do with augmentation. But perhaps those are two different assessments: Can this person think? and What can this person’s thinking accomplish when amplified? We probably need both.

I owe too much to teachers to pretend they don’t matter

This part is personal for me. Next to my parents, teachers and professors have probably been the greatest influence on the person I became. Not an algorithm. Not a textbook. Not a search engine. People.

There are teachers I had decades ago whose impact I can still feel. A professor who pushed back when my answer was too easy. A teacher who made something click that I had convinced myself I simply wasn’t good at. Someone who introduced an idea I probably never would have gone looking for on my own. Someone who expected more from me than I expected from myself.

You don’t just remember what great teachers taught you. You remember how they made you think. You remember how they made you feel about learning. You remember the moment something opened up.

And I think we should be extremely careful not to confuse the arrival of a powerful new technology with the disappearance of everything that got us here. The internet already contains vastly more information than any teacher could communicate in a lifetime. That didn’t make great teachers irrelevant. Because great teaching was never just information transfer.

A great teacher sees that a student memorized the formula but doesn’t understand the concept. They notice when a kid quietly decides, “I’m just bad at math.” They challenge lazy thinking. They create curiosity. They expose students to ideas they would never search for themselves. They know when to explain something differently. They know when to help. And sometimes, critically, they know when not to help. Sometimes the kid needs another ten minutes of frustration. The struggle is doing something.

An AI tutor may eventually become unbelievably good at explaining calculus or chemistry or Shakespeare. Great. But I don’t think that replaces the human being who helps shape the person doing the learning. If anything, it might finally make the teacher’s most important role more visible. Less encyclopedia. More coach. More challenger. More mentor. More person who sees potential before the student sees it. More architect of intellectual development.

That’s not a downgrade of the profession. I think it’s an elevation of it.

We should recognize where we came from while being honest about where we’re going. The tools will change. The assignments will change. The way students demonstrate understanding will change. But the human beings who inspire curiosity, demand more of us, teach us to question ourselves, and help us discover what we’re capable of may become even more valuable in a world where information and intelligence are everywhere.

Zoom out for a second

Today we’re arguing about whether a teenager should use AI to write an essay. That matters. But zoom out far enough and the argument starts to look almost quaint. Because the actual question is much bigger:

How do we educate human beings in a world where cognition itself becomes a technological resource?

That’s not a classroom technology question. That’s a civilizational question.

The Industrial Age changed what people needed to know because machines changed the structure of work. The Information Age changed it again because computers and global networks changed how we found, processed, and distributed information. Now we’re entering something different.

Intelligence is no longer something human beings only possess. Increasingly, intelligence is something we can access. Collaborate with. Direct. Scale. Deploy. And eventually connect to systems that can act in the physical world. That means the educated human being of the future may need a different relationship with intelligence entirely.

The thing I don’t worry about

I’m actually not terribly worried that a child someday won’t be “smarter than AI.” That’s probably the wrong competition. My calculator is already better than me at arithmetic. Google Maps has a better memory for roads than I ever will. A chess engine can annihilate almost every human being who has ever played chess. Fine.

None of those things can tell us what kind of society we should build. And neither can raw intelligence. Capability doesn’t automatically give you purpose. A system may eventually be able to reason across more possibilities than any person alive. It still doesn’t settle: What is worth doing? What is beautiful? What is just? What risks are acceptable? Whose suffering matters? What should we build? What shouldn’t we build? What kind of future do we actually want?

Those are different questions. And oddly enough, the more powerful our tools become, the more consequential those questions become too.

We are not educating children to defeat the machine

Maybe that’s the mental model we need to throw away. The point of education was never really to produce the human being capable of retaining the largest number of facts. Or performing the greatest number of calculations. Those were proxies for capability in a world where intelligence was scarce. That scarcity is starting to change.

The children entering school today may eventually have access to cognitive capability that would have seemed absurdly powerful to previous generations. They may describe an idea and have software built around it. Ask a scientific question and dispatch thousands of simulations. Imagine an object and have machines design and manufacture it. Talk to tutors capable of explaining almost anything in whatever way works best for them. Access specialized intelligence in nearly any discipline almost instantly.

That should make us uncomfortable. It should also fill us with awe.

But none of this means throwing away the educational traditions that built us. I am a product of parents, teachers, professors, and mentors who shaped the way I think long before artificial intelligence entered this conversation. I don’t want to lose that. I don’t think we can afford to lose that.

The challenge is harder than simply replacing an old system with a new one. We have to figure out how to preserve the deeply human parts of education while redesigning the rest for the world our children are actually going to inhabit.

Because the great educational challenge of the next generation may not be teaching kids how to compete with artificial intelligence. It may be teaching them what is worth doing with intelligence this powerful. What should they ask? What should they create? What should they question? What should they believe? What should they refuse to build? How should they decide when the machine is wrong? What responsibilities come with this kind of capability? And what kind of future is worth bringing into existence?

Machines may become extraordinarily good at answering questions. Humanity is still responsible for deciding which questions matter.

And maybe that’s what the best teachers have been trying to teach us all along.