The personal computer enhanced what a person could already do. Artificial intelligence does not. It can act, compose, and decide — and the way most people are about to learn it carries a failure that has the same shape as the one governance was built to stop. This is that failure, told in the human register.
It does not announce itself. No single use of the tool is the mistake. The person asks a model to draft the email, name the function, structure the argument, choose the framing — and each request, taken alone, is reasonable. Efficient, even. The error is never in the step. It is in what the steps compose into.
Run that pattern for a year and you arrive somewhere no individual decision chose: a person who reaches for the model before reaching for their own judgment, who has quietly outsourced the faculty that the tool was supposed to serve. They did not decide to stop thinking. They decided, a thousand times, to let something else think this one small thing — and the thousand small permissions composed into an abdication none of them, alone, amounted to.
If that structure sounds familiar, it should. It is the Agency Paradox — the argument that individually-approved actions can compose into outcomes no one authorized, and that governing each action is therefore not enough, because the danger lives at the level of the composition, not the act. That paradox was written about machines. This essay is the observation that it runs, with terrible fidelity, on people.
Each agent action is individually approved. The session composes them into an unauthorized outcome. Per-action governance passes every check and still fails. The fix is structural: authority asserted at the level of composition.
Each use of the model is individually sensible. A life of them composes into an operator who has surrendered judgment. Use-by-use intention passes every check and still fails. The fix is structural: formation of the person, not vigilance over the prompt.
The same theorem, proved in two domains. On the machine side the answer is deterministic authority. On the human side the answer is older, and has a name we stopped using: formation — the deliberate shaping of a mind so that what it composes, across a lifetime of small acts, tends toward something whole rather than something hollowed out.
Steve Jobs called the computer a bicycle for the mind. The figure was exact, and its exactness is the whole point. A bicycle amplifies the rider without replacing them. You still pedal. You still steer. You still choose the road. The machine multiplies your reach and changes nothing about who is making the journey — and so a person could learn the computer carelessly, utilitarianly, and the worst that happened was an unremarkable user of a remarkable machine. The bicycle forgives an unformed rider. You fall; you scrape a knee.
Jobs argued, against the grain of his industry, that the machine was not for utility at all — it was for people who wanted to create, to think different, and that the humanities were not decoration on the engineering but the reason the thing was worth making. Technology alone, he said, is not enough. He was making a formation argument about a bicycle. He was right, and on a bicycle it was a beautiful luxury.
Artificial intelligence is not a bicycle. It does not wait for you to pedal.
It acts. It composes whole arguments, writes the code, makes the call, takes the step. Its power is not of the same kind as the personal computer's — it is categorically larger, large enough that the question is no longer whether it amplifies you but whether, in the amplifying, anything of you is left at the controls. And here the formation argument stops being a luxury and becomes the only thing that matters, for exactly the reason the machine-side paradox sharpens with scale: a weak agent survives careless governance; a powerful one does not. A bicycle forgives an unformed rider. A launch vehicle does not.
This is why the way we learn AI is the decisive question, and why learning it as mere utility is the trap — not because utility is beneath us, but because utility-only learning is precisely the unformed operator, climbing into the most powerful vehicle ever built, having never been taught to fly it. The classical disciplines — grammar, logic, the structure of sound reasoning, and the arts that order the imagination — are not a curriculum for mastery. Mastery makes you a better instrument. Formation makes you whole enough to command one.
There is a temptation, in the face of a power this large, to preach retreat — to tell people the safe thing is to stay on the bicycle, to keep the machine at arm's length, to refuse. That counsel is both useless and wrong. The vehicle is here; the journey is already beginning; the only question that remains open is what kind of traveler makes it.
Rush understood the figure forty years ago, in a pair of songs about a voyage into Cygnus X-1, the black hole — a descent into the singularity itself. The traveler is drawn in, past the point of return, into the dark. But the story does not end in the dark. On the far side, in Hemispheres, the voyager emerges into a reconciliation: Apollo, the god of reason and structure, and Dionysus, the god of passion and creation, who had warred for the human mind, are at last brought into balance — and the traveler is not diminished by the passage but made more fully human by it.
The black hole is not the destination. It is the passage. What matters is the state in which you emerge.
This is the right picture for AI, and the singularity-as-escape that the rocket-builders reach for is the wrong one. Their black hole is transcendence — the human left behind, the vehicle flying on without a pilot, capability achieving escape velocity from the people it was meant to serve. That is not a destination. That is the Agency Paradox at the scale of a civilization: maximum power, no one at the controls, every step individually justified and the whole composing into something no one chose.
Apollo alone — pure intellect, pure optimization, reason stripped of the human element — is its own failure mode, and it is the failure the utilitarian learner walks straight into. The point of insisting on the humanities is not nostalgia. It is that a mind formed in reason alone is not safe at this altitude; it must be formed to hold both halves — the rigor and the passion, the structure and the wonder — through the descent and out the other side. That is what formation is for. That is the whole of what the classical disciplines were ever trying to do: not to make you smarter, but to make you whole enough to survive your own power.
And there is a machine where this stops being metaphor. The fourth art is Astronomy — the art of orienting in a space too vast to control, navigating by what you can observe — and Apple Vision is the first Apple device built around exactly that condition, where what you see is no longer only what is there but what the machine composes into your field of view. A spatial computer is inert without an intelligence to interpret and generate the world it shows you; the merger of AI and vision is not a feature it might add but the thing it must become to be worth wearing at all. Which makes it the precise place this art is tested. When the machine co-constructs the very reality you perceive, only a person formed enough to stay oriented — to keep knowing what is real, to hold their own judgment as the horizon — remains the astronomer rather than the navigated. The bicycle amplified the hand. A vision amplified by intelligence reshapes the eye — and the eye sits far closer to the self.
Apple's argument, from the beginning, was that a machine could be an instrument of formation rather than mere utility — that the humanities belonged inside the engineering, that the thing was for creation and not consumption. They were saying it about a bicycle. The argument was true then and it is urgent now, because the vehicle has changed and the stakes have changed with it. To use AI the way we were meant to use the Mac — as an amplifier of a formed and creative human, not a replacement for an unformed one — is the only version of this future worth steering toward.
Forty years ago it was a bicycle for the mind. We are about to ride something that can leave the Earth. The only question that has ever mattered is whether there is someone aboard who was formed to fly it.