The question no one is asking about AI.

Every conversation about these machines is a conversation about the machines.

What can it do now. What will it be able to do next year. Which jobs go. Which industries survive. Whether the thing is genuinely intelligent or an extraordinarily good imitation, and whether that distinction will matter to anyone once the invoices start arriving either way.

All reasonable questions. I have asked most of them, some of them out loud in rooms where I should probably have been quieter.

But they are all questions about the tool. And the tool is not the interesting part.

The question almost nobody is asking is this one: what happens to a person who never has to do a hard thing badly first?

How judgment was always manufactured

Consider how anyone has ever become good at anything.

You attempt the thing. It is poor. You cannot fully see why it is poor, but you can feel the gap between what you meant and what came out, and the gap embarrasses you. You do it again. Somewhere around the fortieth attempt you begin to notice, in advance, the decision that will make the next one bad — and you avoid it before you make it. That noticing is what we call judgment, or taste, or instinct depending on the field. It is the only asset in any creative business that cannot be bought.

Now look closely at how it was produced. Nobody ever set out to acquire it. There is no course. It arrived as a by-product — the residue of hundreds of unglamorous attempts that were mostly failures and mostly wasted, in the narrow sense that almost none of them shipped.

That waste was not waste. That was the machinery.

And that machinery is now being switched off, quietly, by people making individually correct decisions.

Because here is the trap, and it is a good one. Every single time, using the machine is the right call. This afternoon, on this task, with this deadline, the tool produces something better than my first attempt would have been, in a fraction of the time, and I would be a fool and a bad businessman to refuse it. There is no instance in which the disciplined choice looks smart. The cost never shows up in the decision. It shows up ten years later, in a person who can produce anything and judge nothing — who can generate forty options and has no idea which one is good, because the faculty that would have told them was built out of the very attempts they were sensible enough to skip.

Compound interest is the great argument for patience. Nobody mentions that it also runs in reverse, and that the reverse feels identical to efficiency while it is happening.

What I hand over without hesitation

I want to be clear that I am not making a romantic argument. Refusing good tools is not a virtue, it is vanity dressed as principle, and every craftsman in history who took that position has lost to someone who did not.

So I use these tools daily, at close range, in real work. I have no interest in doing by hand what a machine does better.

I hand over the middle of the work gladly. The production. The variation — the fiftieth version of a layout, the twentieth way of phrasing a paragraph. Research synthesis, the reading of forty documents to find the three that matter. The scaffolding, the boilerplate, the structure, the first pass at anything where a first pass is simply an obstacle between me and the second. This site you are reading was built that way and I am not going to pretend otherwise; the thinking in these essays is mine, some of the drafting is not.

That kind of leverage is not a compromise. It is the whole reason a small operation on this continent can now hold a standard that used to require a building full of people, which is a genuinely new fact about the world and one of the few unambiguously good ones.

The question is not whether to use it. The question is what you deliberately keep, knowing it would be faster not to.

What I keep by hand, and why

Deciding what should exist. This is the founding question and it does not delegate. A machine can tell you what has worked, because it has read everything that worked. It cannot tell you what should exist and does not yet, because that judgment is made of conviction, and conviction is made of caring about something in particular. Every serious company begins with somebody who wanted a specific thing to be true in the world. Nothing about that has been automated, and I do not believe it will be.

First contact with reality. The customer conversation. The pitch that goes badly. The no. There is enormous temptation to run all of this through a layer — to have the awkward exchange summarised, the objection pre-answered, the rejection softened into a bullet point. Do not. The information in a no is not in its content. It is in the pause before it, the thing they did not say, the discomfort you feel afterwards that you cannot yet articulate. That discomfort is data of a kind no summary will ever carry, and it is where every good pivot I have ever made actually came from.

The last look. Whatever the tools produce, somebody has to stand in front of the finished thing and decide whether it is good — not whether it is competent, whether it is good, which is a different question with no metric attached to it. That decision is the entire value of a company with taste, and it is unimprovable by the very system that produced the work, because the system is optimising for resemblance and the standard is not.

And some difficulty on purpose. This is the one I have had to become deliberate about, because it has no business justification whatsoever.

Occasionally I do the hard version anyway. Sit with the blank page. Work the problem out longhand. Not because the result will be better — it usually is not — but because the faculty I am using is the one I intend to still have in ten years, and faculties are maintained the way muscles are. For the whole of human history, difficulty was compulsory. It was simply the price of the outcome, and it built us as a side effect while we were busy trying to avoid it.

That is over. The difficulty is now optional. Which means, for the first time, it has to be chosen — and almost nobody will choose it, because on any given afternoon there is no reason to.

What I am actually afraid of

Not the jobs. The jobs are real and serious and I do not wave them away, but that argument has been had, and every previous version of it eventually resolved into a different distribution of work.

What I am afraid of is a generation that outsources judgment before it has any.

Think about where this lands hardest. Somebody twenty years into a career uses these tools as an amplifier — they already have the eye, the machine just removes the drudgery between them and the output, and they get better. But somebody at the beginning has nothing to amplify. They are being handed, at the exact moment they should be doing their two hundred bad attempts, a machine that makes bad attempts unnecessary. They will produce far more impressive work than my generation did at their age. They will not, unless they choose to, develop the ability to tell whether any of it is good.

That is not a failure of character. It is a failure of design. We have built an environment where the developmental path is optional and the shortcut is always right there, and then we will blame the young for taking it.

I think about this most in relation to where I build. I have argued elsewhere that the great opportunity for this continent is that the tools finally arrived here at the same price they arrived everywhere else — that a young person in Pretoria or Lagos or Nairobi now has the same instrument on the desk as anyone in California. I believe that completely.

But there is a specific way to waste it, and it is the oldest pattern we have. We have spent a century exporting the raw thing and importing the finished thing — making what we do not consume, consuming what we do not make. The new version is subtler and moving fast: adopting the output of intelligence we do not build, on infrastructure we do not own, trained on data we hand over without negotiating, to produce work whose real margin accrues somewhere else entirely. The interface changes. The trade does not.

Using the tools is right. Building only on top of them, and never underneath, is the same arrangement in a better font. Capability is the thing to accumulate — in people first, in institutions after. That accumulation still happens the slow way, and it still requires a generation willing to be bad at things in public for a while.

The part that is not about work

There is a version of this argument that is entirely economic, and I could have stopped at the last section and it would have held.

But I do not think a person is only what they produce. Struggle is not an inefficiency in the human design that we have finally engineered around. Most of what I value in the people I admire — patience, discernment, steadiness, the particular calm of someone who has been through something — was not installed. It was earned, in conditions they would have escaped if escape had been available.

We are the first generation with a general-purpose escape from difficulty. That is an astonishing gift and I use it every day. It also means that whatever we would have become through struggle, we now have to become on purpose, or not at all.

So the question is not what these machines can do. They can do a great deal, and next year they will do more, and orienting your life around that race is a way of never asking anything of your own.

The question is: what am I unwilling to stop being able to do?

Answer it early. Write the list down while you still have the abilities on it. Then defend them like an asset, because that is exactly what they are — the last ones that will not be commoditised, held by the shrinking number of people who decided, for no good reason on any particular afternoon, to do the hard thing by hand.