I Built a Hut With My 10-Year-Old Nephew. It Taught Me More About Cognition Than a Decade of Essays

A few weekends ago, my ten-year-old nephew and I built a hut in the woods.

“Built” might be generous. We had salvaged planks, bent nails, a hammer, and the kind of plan that feels solid until wood, slope, weather, and gravity join the conversation. By the end, we had something crooked, usable, and slightly absurd. It may survive the winter. I would not put money on it.

What stayed with me was not the hut itself. It was the quality of attention the work demanded, and how different that felt from the attention I have spent years training in front of screens. While we were wrestling beams into place, I kept thinking about all the essays, reports, briefs, and tidy arguments I have written over the years. Then the uncomfortable thought arrived: what if two days in sawdust trained parts of the mind that ten years of polished writing barely touched?

That question matters more now because writing has changed status. For a long time, producing text was treated as evidence of thinking. Now machines can draft, summarize, reorganize, and imitate argument with unnerving fluency. If the old exercises can be automated, we should ask whether they were the best cognitive training we had, or simply the easiest to administer at scale.

The world gives cleaner feedback than a teacher can

When you build a hut, reality grades every decision immediately.

If a span is too wide, the board sags. If your base is off level, the structure starts leaning before you are emotionally prepared to admit it. If you forget where the wind actually comes from, the roof reminds you later. Nothing is symbolic. The correction is built into the task.

That makes the learning strangely pure. You do not have to infer whether your reasoning worked. The structure tells you. There is no rubric translating the outcome into institutional language. The wood either carries the load or it does not.

Compare that with the typical essay. You submit it. Days pass, sometimes weeks. Then a grade appears with comments in the margins: underdeveloped point, weak evidence, unclear transition, argument not fully convincing. Some of that feedback is useful. Much of it is delayed, interpretive, and filtered through another person’s preferences. You can improve from it, but the loop is slow and fuzzy.

The difference is not just speed. It is the type of truth each activity exposes. In writing, it is possible to sound coherent before you are coherent. In building, the pretense does not survive contact with the beam.

Embodied thought is still thought

We often talk as if “real thinking” happens in language, somewhere behind the eyes, and everything else is implementation. That view flatters people who are good with words. It also misses how minds actually work.

Cognition is not only symbol manipulation. It is also perception, timing, proprioception, spatial judgment, physical prediction, sensory integration, and adaptation under constraint. The body is not a delivery system for the brain. It is part of the process by which the brain learns what the world is like.

That became obvious over two days in the woods. We were not discussing geometry as a school subject, but geometry was everywhere. Angles were no longer lines on paper. They were the difference between a brace that held and one that twisted. Weight was not a number. It lived in the hands, in the awkward shift of a board that looked manageable until you tried to hold it steady and hammer at the same time. Planning was not an abstract sequence. It kept being revised by the hill, the trees, the length of the planks, and our own fatigue.

This is what people mean when they talk about embodied cognition, though the phrase can sound more exotic than the idea really is. Thought changes when the body is involved in a meaningful way. The environment becomes part of the reasoning system. A sketch, a tool, a half-built frame, even the resistance of a material can do cognitive work for you. They are not props. They are active partners in the process.

My nephew understood this faster than I did. Children often do. He would test a possibility, feel it fail, and update instantly. I was more tempted to defend the original plan, as if commitment could persuade the wood to cooperate. The child was closer to scientific method. I was closer to office culture.

Essays reward a useful skill, then overextend it

I am not arguing that writing is worthless. Writing matters because language clarifies thought, exposes contradiction, and forces some degree of order onto vague intuitions. If you cannot explain an idea, there is a decent chance you do not understand it well. There is a reason serious disciplines still depend on prose.

But we have confused an important skill with the master skill.

Modern education elevated abstract cognition for understandable reasons. Reading, writing, and testable reasoning scale well. They fit classrooms, timetables, exams, and bureaucracy. Thirty essays can be assigned and compared. Thirty construction projects require space, materials, supervision, risk tolerance, and a system willing to accept messy outcomes. Institutions tend to prefer what can be standardized over what changes people most deeply.

That preference has consequences. Students learn early that success often means producing the right shape of answer. They learn to imitate understanding before they have it. They learn to optimize for evaluation. Plenty of bright people become excellent at sounding like they know.

The essay is almost designed for this kind of performance. You can restate familiar points with elegant transitions and earn respectable marks. You can borrow the cadence of insight without living through the friction that gives insight weight. And now, of course, you can outsource much of the surface work to a model trained on the internet’s vast archive of plausible sentences. The machine is very good at the outer shell.

A hut is rude in a healthier way. If you misunderstand leverage, support, or sequencing, the outcome degrades immediately. The world does not care that your intentions were nuanced.

Physical work trains a different relationship to error

The most striking part of building with a child was how normal failure felt. We changed our plan constantly. Not because the initial plan was foolish, but because every real project is a negotiation with conditions you do not control.

The trees were farther apart than we wanted. The ground was less even than it looked. One board split where we needed strength. Another was shorter than memory had promised. We improvised a brace, abandoned one wall idea, lowered the roofline, and rethought the entrance. None of this was a crisis. It was simply the work.

That style of learning is underrated. School often treats error as a sign that you were insufficiently prepared. The physical world treats error as information. If you are paying attention, every failed attempt sharpens the next move. The loop is tight. You test, perceive, adjust, repeat.

This matters because many adult jobs have drifted toward environments where feedback is delayed and ambiguous. You can spend months inside slide decks, strategy memos, planning documents, and performance language without ever discovering whether the underlying model fits reality. Whole careers become exercises in highly literate drift.

Building anything tangible interrupts that drift. It restores a habit of contact. You stop asking whether the argument sounds right and start asking whether it works. Those are different muscles.

The bluffing problem just got larger

For years, people treated writing as both the medium of knowledge and proof of effort. If a student turned in a competent essay, we assumed they had read, processed, and synthesized. The assumption was never fully safe, but it was workable enough.

Generative AI has blown a hole in that bargain.

Now someone can produce text that looks organized, informed, and reflective with very little internal transformation having taken place. Teachers know this. Employers know this. Students know it most clearly of all. Once a task becomes easy to simulate, it loses value as a signal of understanding.

Some responses to this shift are defensive. Ban the tools. Return to handwritten exams. Search for traces of authenticity in style. Some of that may be necessary in specific settings. But the deeper opportunity is elsewhere. If symbolic production is becoming cheap, we should stop worshipping it as the default path to mental development.

That does not mean abandoning abstraction. It means placing it in a healthier relationship with reality. Let the machine help with summaries, drafts, references, and first-pass structure. Then spend more human time where friction lives: prototyping, repairing, observing, building, experimenting, negotiating with material limits, and learning from consequences that cannot be faked.

Education was optimized for administration, not cognition

This is the part people resist, because it sounds like a complaint about schools when it is really a complaint about systems.

Mass education had to solve a logistical problem. It needed formats that could be replicated across large populations with limited resources and relatively consistent assessment. Abstract tasks fit that requirement beautifully. Essays, worksheets, exams, and lectures are efficient containers. They are easier to distribute than wood, tools, lab access, mentors, workshop space, and open-ended projects.

The result is a culture that quietly equates what is easy to test with what is worth knowing.

You can see the distortion everywhere. A child who writes a clean essay on ecosystems is judged intelligent. A child who can repair a bike, improvise a shelter, diagnose why water is pooling near a foundation, or understand load distribution by feel is often treated as “practical,” which is a subtle demotion in many elite settings. We say we value problem solving, but often mean verbal performance about problems.

This hierarchy made a certain kind of sense when paperwork and formal writing were scarce, expensive, and socially decisive. It makes less sense in a world where machines can generate plausible paperwork in seconds.

A school system that took this seriously would not stop teaching writing. It would reduce the amount of human time spent on low-yield symbolic repetition and increase the time spent making ideas answer to reality. Students would still read, argue, and explain. They would also build, measure, test, grow, repair, simulate, and revise based on what actually happens.

The same applies outside school. Many professionals spend absurd amounts of energy producing documents whose main function is to satisfy process. If AI can absorb more of that load, the gain should not simply be more documents at higher speed. It should be more contact with the world those documents claim to describe.

What writing still does that wood cannot

There is a temptation here to swing too far and declare physical making the superior form of intelligence. That would be its own mistake.

Writing remains one of the best tools for reflection across time. It lets you compress experience, compare cases, preserve reasoning, and communicate beyond the people standing next to you. You cannot build public knowledge from tacit skill alone. Civilization still needs people who can think in language, not just with their hands.

And some important domains are abstract by nature. Law, philosophy, mathematics, policy, and history do not become better merely because we add sawdust. There are forms of rigor that only words and symbols can support. A society that stops training those capacities would become shallow in a different direction.

The issue is proportion. We built a prestige economy around disembodied fluency and then called it intelligence. AI is exposing how much of that fluency was pattern production all along. This should push us to ask a sharper question: which human capabilities become more valuable precisely because they are anchored in experience, judgment, and material consequence?

A child who has written five elegant pages about resilience may understand the concept. A child who has spent an afternoon trying to keep a crooked wall upright while rain threatens and the nails keep bending has learned something denser. Ideally, you want both. For a long time, we pretended the first could substitute for the second.

A better division of labor is available now

The phrase that keeps circling in my head is simple: use machines for abstract repetition, and reserve more human energy for concrete creation.

That is not a manifesto for everyone to become a carpenter. The relevant activities are broader than woodworking. They include running experiments, debugging hardware, cooking, planting, repairing, prototyping products, staging performances, designing spaces, building community infrastructure, and any task where the world pushes back in ways that sharpen judgment.

The common feature is contact. You are not merely arranging symbols about reality. You are acting inside it and being corrected by it.

Once you see the difference, a lot of modern dysfunction becomes easier to understand. Organizations produce endless language because language is easy to copy, circulate, and reward. Yet the things that change outcomes often come from the slower layer where people try something, observe what breaks, and adapt. Those people may write fewer memos. They usually know more.

My nephew did not need a theory of cognition to benefit from the weekend. He needed a hammer, some bad materials, a problem worth solving, and enough freedom to discover that ideas feel different when the world is allowed to answer back. I needed the same thing, though it was mildly embarrassing to realize it this late.

The hut was crooked, and the lesson was clear

We tend to imagine intelligence as something that happens at a desk and then gets applied elsewhere. Building that hut reminded me that for many kinds of understanding, the order runs the other way. You do not think first and then meet reality. Often you meet reality, fail a bit, adjust, and only then earn a thought sturdy enough to keep.

This is why the rise of generative AI should not push us into panic about human irrelevance. It should push us into specificity. If machines are increasingly competent at producing the symbolic residue of thought, then the distinctively human work shifts toward judgment grounded in context, skilled attention, physical improvisation, and experience that cannot be downloaded as style.

That applies to education, but also to adulthood. Plenty of us have spent years mistaking verbal polish for development. Then a simple physical task exposes the gap. A wall that will not stand can be a better teacher than a hundred pages of competent prose because it refuses to flatter you.

The hut in the woods was unstable, patched together, and probably a poor investment if measured by square footage. As a lesson about how minds learn, it was unusually solid.