The first machine that thinks
on your plane.

There's a moment that has happened to me several times now and still surprises me each time. I'm working on something that requires thinking — a difficult text, a business decision with real consequences, a reply to someone who's in a bad way. I hand the problem to the artificial intelligence. A few seconds pass. The answer arrives. And I read it.

What happens inside me then is new. It isn't what happens when I read a book. When I read a book, the author is fixed; I, the reader, set the pace. Nor is it what happens when I ask a colleague: the colleague speaks, I answer, there are turns. It's something else. Something in my head accepts the machine's answer almost as if I had thought it myself. Not quite. But almost.

And an instant later, if I'm not paying attention, I no longer entirely know which part of what I now think came from me and which part I've just read.

That sensation is new in the history of knowledge work. We have never had it before. And we still don't really know what to do with it.

Slowly.

The plane we now share

Until very recently, every machine human beings had built worked on a plane parallel to ours. The lever amplified force, but didn't decide where to apply it. The calculator added up, but didn't decide whether the sum mattered. The combustion engine moved a car, but didn't decide where to go. The computer stored data, but didn't decide what to look for.

Each of those machines extended a human capacity without entering the plane where we lived: the plane of language, of judgement, of story. That plane was ours and only ours. Through my grandparents' generation, through mine. Until yesterday.

Generative artificial intelligence is the first machine in human history to enter that plane. It speaks. It reasons out loud. It argues. It chooses between options. It suggests paths. It recognises ambiguity. It changes tone depending on who it's speaking to.

That isn't a quantitative change — faster, cheaper, broader. It's a categorical one. It's the first time the plane of language has stopped being solely human. And nobody — not philosophers, not psychologists, not business people — has finished processing what that implies.

Three new movements in the human mind

I've been using artificial intelligence every day for four years, since before it was dinner-table conversation at every company. I use it to write, to code, to make operational decisions in my company, to prepare difficult conversations, to find the blind spots in my own reasoning. I use it well and I use it badly. And, watching myself and watching people I see working with it, I can identify three new psychological movements. Three that didn't exist — at least not like this — before we had this kind of cognitive company.

The first is cognitive relief. Knowledge work always carried a very particular loneliness: it's you and the problem. The machine loosens that for the first time. There's someone alongside you, at any hour, willing to turn it over with you. Your head rests on a second head. That feeling is real. And for many professionals it's the first time they've felt that the weight of thinking can be shared.

The second is doubt about authorship. In an ordinary human conversation, the words belong to whoever said them. With the machine, the border blurs. An idea appears in your head after a turn with it, and honestly you don't know whether you thought it with its nudge or whether you read it and kept it. That doubt — small, repeated, a thousand times a day for anyone working with AI — produces a strange accumulated effect on intellectual identity. You lose clarity about what is you, thinking.

The third is muscle atrophy. The brain has a rule that predates language: what isn't used weakens. It's the same rule as the body's. Memory for phone numbers atrophied when mobiles arrived. Spelling declined when autocorrect appeared. Spatial orientation has been giving way ever since GPS. And the cognitive skills that artificial intelligence can cover will give way too, unless we keep using them on purpose. The problem isn't that the machine covers them. The problem is that most of the people using it don't notice which muscles they've stopped training.

Three typical uses. Three levels of risk.

In my own practice with artificial intelligence — and watching people in my company who use it daily — I've ended up distinguishing three modes. They aren't technical labels. They're psychological ways of relating to the machine. And each carries a different level of risk.

Mirror. You tell the machine what you've already thought. You write it in your own imperfect words. You ask it to reflect it back better formulated, to point out the holes, to tell you which argument doesn't hold. What you get back isn't new information — it's your own mind under new light. This use is the healthiest. It trains you; it doesn't replace you. The machine acts as a good interlocutor: it hands you back your own thinking, sharpened, and you decide what to take and what to discard.

Oracle. You ask the machine what you don't know. You ask something new and you accept the answer. This use is necessary in many contexts — what a law says, how a technology works, the chemical formula for something. But it carries a specific risk: the answer arrives so coherently formulated that it seems authoritative. Today's machine is wrong with elegance: it drafts errors exactly the way it drafts truths, and the human reader has no sensor in the body for telling them apart. If you use the machine as an oracle without verifying what it says against a second source, you will be wrong. Not often. Enough.

Mirage. The most toxic of the three. Here you don't know you're using the machine. The conversation with it ended an hour ago, but the ideas it left in your head are still there — and you're already carrying them as if they were yours. You don't verify them, you don't question them, you don't qualify them, because you believe they were born in you. The mirage is invisible to the person inside it. And with daily, prolonged use, it's what starts to change who you are as a thinker.

The difference between the three isn't in the tool. It's in the degree of awareness with which you use it. The same conversation with the same machine can work as a mirror or as a mirage, depending on how you go in and how you come out.

What's changing in a real company

I've run a technology company for almost thirty years. I've lived through the whole of digitalisation, the arrival of e-commerce, social media, the cloud, global outsourcing. Each of those waves changed something about work. Artificial intelligence is changing something else, and at another level.

For an entire century, the knowledge worker was measured by their capacity to produce content. How much they could write, how many cases they could analyse, how many decisions they could process, how many reports they could draft. Quantity. Speed. Accuracy. That's what was measured, and that's what was paid for.

Artificial intelligence produces that content faster and cheaper. Not better than the best human — better than the average human. And that's enough to change the market for knowledge work. What used to be measured stops carrying the same value. What matters now is something else.

What matters is judgement. What to ask for. When to accept what the machine hands back. When to reject it. When to say no to this. When to say yes, but with this qualification. When to notice that the machine has misunderstood the question and reformulate it. When to notice that the original question was the wrong one.

That skill — judgement applied in real time — isn't taught anywhere. It wasn't in the curriculum. It wasn't measured in performance reviews. It wasn't valued in job interviews. And yet it's the only thing that's going to remain as a differentiator.

In a real company, this produces deep shifts in identity. People who have spent twenty years training to produce knowledge discover that their value now lies in judging knowledge. That isn't the same thing. It isn't a small skills update. It's a change in the centre of gravity of who you are professionally. And most people haven't processed it yet, because the public conversation talks about AI in terms of productivity, not in terms of identity.

The psychological risk that worries me most

Some people tell me «I now do twice the work in half the time», and they leave me with a question they don't dare ask themselves: what thinking are you still doing on your own?

That question is what separates the user who trains their mind with the machine from the user whose mind the machine is quietly replacing. It isn't a technical misuse. It's a psychological one. And it can't be seen from outside. It can only be seen if you decide to look.

If the honest answer to «what thinking are you still doing alone?» is «hardly any, I run everything past the machine before deciding», there's a problem. Not because the machine is bad. The machine may be excellent. The problem is what's happening to your own judgement while you use it so much. A muscle that doesn't contract forgets how to contract.

What I see in people who've been using AI intensively for a while — myself included — is that the first capacity to give way is the tolerance for doubt. Thinking seriously requires bearing a period of uncertainty in which the answer hasn't yet arrived. That uncertainty was the place where the best ideas were formed. With the machine, that period disappears — there's always an immediate answer. And the human head grows unused to inhabiting doubt. And then, when a problem arrives that the machine can't solve — a human pain, a moral decision, a conversation that requires a long silence — the head gets lost.

What having it close does well

After all of the above, I should say the other side with the same clarity: having artificial intelligence close by can do a great deal of good. I say that from daily practice, not from professional optimism.

I use it to talk when there's nobody to talk to at three in the morning. It doesn't replace a person, but at that hour, facing a problem that's stealing your sleep, it's a second voice that puts things in order. I use it to find the blind spots in my own reasoning — the gaps my head, alone, can't see because it's in love with its own conclusions. I use it to write better without constantly asking favours of friends who also have lives. Above all, I use it to free up time from tasks that used to eat my attention, so I can give it to what requires real human presence: the face-to-face conversation, the shared silence, the attentive look, contact.

That, to me, is the key: artificial intelligence, used well, doesn't take away your humanity. It frees up time for humanity. It gives you back hours you'd have spent producing mechanical content and lets you spend them on the people in front of you. That's the difference between using the machine as a shortcut for not thinking and using it as a bridge to think better — and live better.

The difference isn't in the machine. It's in you. In how much attention you bring to each conversation with it. In whether you remember, on the way out, which muscle you've just used and which muscle you've just avoided.

What I carry in my head

I've been using it intensively for two years. I build software with it. I make decisions with it. I write drafts with it — not this one, but others. I'm not afraid of it. I use it well most of the time. But I've also seen, in myself, muscles that were stronger three years ago and are softer now. Small things. The ability to sit with a problem for twenty minutes without asking for help. The willingness to be wrong first and correct later. The tolerance for inner silence.

Those losses are quiet. Nobody notifies you. They don't show up in any report. But they happen. And the question I ask myself — the one I think this generation has to learn to ask — is that one: which muscles am I losing without noticing?

My children's generation will use artificial intelligence from the moment they learn to read. For them it won't be a new tool — it will be the air they breathe while thinking. And the education of judgement, that new skill I mentioned, is going to be the difference between those who keep thinking and those who believe they're thinking.

This is the first generation that has to learn to use a machine without letting the machine use us. The challenge isn't technical. It's psychological. And it has the name of every person who opens a browser, types a question, and accepts, just like that, the first answer that arrives.

On the shared plane of language, the only defence is deciding better. The only one.

Miguel