The word for what's happening

There is a moment I keep returning to. It is late, the room is quiet, and I am building a website with a machine that writes most of the code while I watch. I described this in another essay, the strange joy of finally making something I had imagined for years, with a tool that does the parts I never learned. That essay was about agency, about what it means to build something yourself when most of the building is done by something else.

This one starts where that one left off. With a feeling I did not have a word for.

Because somewhere in that night, between one working page and the next, something tightened. Not doubt exactly. Not fear. A small physical unease, the kind you notice in your shoulders before you notice it in your thoughts. The work was going well. That was the problem. It was going well in a way that did not feel like mine, and I had no language for the difference.

I want to take that feeling seriously. Not explain it away, not dramatize it. Just look at it, because I think a great many people are feeling some version of it right now, and because the reason it stays nameless is the most interesting thing about it.

A word that does most of the work

Start with the obvious thing nobody questions. We call it artificial intelligence.

The phrase is so settled that it feels like a description rather than a choice. But sit with it for a second. What the machine does when it helps me build, or drafts a paragraph, or answers a question, is extraordinary pattern recognition across more text than any human could read in a thousand lifetimes. It is cognition of a kind. Whether it is intelligence, in the sense we usually mean when we use that word about a person, is a genuinely open question, and the people who build these systems are the ones asking it most loudly.

Yann LeCun, who won computing's highest honor for his work on the neural networks underneath all of this, has said plainly that today's language models are "an off ramp on the road to human-level AI," a detour rather than the path. He likes to point out that

a house cat understands the physical world, can plan complex actions, and can reason better than the biggest language models.

A cat. The comparison is not an insult to the technology. It is a precise observation about what intelligence, the embodied animal kind, actually involves, and how much of it has nothing to do with language at all.

Melanie Mitchell, one of the clearest thinkers on this, names the underlying mistake directly. One of the field's recurring errors, she argues, is the assumption that intelligence can be cleanly separated from the body, that it is a kind of disembodied information processing you could lift out of a brain and run anywhere. Human thinking does not work that way. It is tangled up with the whole nervous system, with sensation, with the fact of having a body that wants things and fears things and gets tired.

I am not raising this to settle whether machines think. I am raising it because the word we reach for, intelligence, quietly imports a whole set of assumptions, and those assumptions shape what we can notice. Which is, as it happens, a very old idea.

The limits of language

A hundred years ago, a young Austrian wrote a sentence that has never stopped being quoted, usually badly.

"The limits of my language mean the limits of my world." — Ludwig Wittgenstein, Tractatus Logico-Philosophicus

People treat the line as a tidy slogan about vocabulary, learn more words, see more world. Wittgenstein meant something stranger and deeper. He was not talking about how many words you know. He was talking about how the structure of what you can say sets the outer edge of what you can think clearly at all. The self, in this picture, is not an object inside the world but the boundary of it, the way the eye is the edge of the visual field and never appears inside it.

What lies past that edge does not announce itself as a missing word. It shows up as silence. Or, and this is the part I care about, it shows up in the body, as the thing you feel before you can say it. The famous last line of the book is "whereof one cannot speak, thereof one must be silent." But silence is not nothing. Anyone who has sat with an unnameable feeling knows that silence has a texture, and sometimes a pulse.

Here is the turn that makes Wittgenstein worth keeping rather than quoting. He spent the second half of his life arguing with the first half. The young man who drew a hard line around language became an older one who decided meaning was not a picture at all, but a use, a move in a game played between people.

Words mean what we do with them.

The thinker who once fixed the limits of the world spent his later years showing that those limits move, that they are made and remade in the practice of speaking.

I find that enormously consoling. It means the namelessness is not a wall. It is a frontier.

Not a better tool, a different world

So what is actually happening when a technology arrives and we cannot quite describe our own experience of it?

Thomas Kuhn gave us the word everyone now overuses. Paradigm shift. He was writing about science, about how a settled way of seeing the world gives way, not gradually but in a lurch, to a new one. The thing worth recovering from Kuhn, under all the management-seminar varnish, is how unsettling he understood the process to be.

In a genuine shift, practitioners do not simply learn new facts. They begin to see different things when they look at the same place.

And crucially, the new way is not just the old way plus improvements. Kuhn pointed out that when Copernicus put the sun at the center, his model was not, at first, better at predicting where the planets would be. It was not a superior tool. It was a different picture, and people resisted it precisely because the old picture still worked perfectly well for getting things done.

This is the distinction I keep wanting to make to people who ask me whether AI is just the next spreadsheet, the next search engine, the next thing we will absorb without much fuss. Maybe. But a tool is something you pick up to do a task you already understand. A paradigm shift is when the task itself, and your place in it, quietly changes shape while you are still using the old words for it.

The unease I felt that night was, I think, the gap between the two. I was using tool language, I am building a website, while something paradigm-shaped was happening underneath — the very meaning of building, of authorship, of skill, sliding around beneath a sentence that had not changed.

What the body knew first

Here is where my own two disciplines, the linguistics and the psychology, stop being separate subjects and become the same observation.

The neuroscientist António Damásio spent his career on a single unfashionable idea, that emotion is not the enemy of reason but part of its machinery.

Cognition, in this view, is not the cool processing of a world held at arm's length. It is enacted, brought forth, through a body acting in an environment.

Damásio studied people who, through specific brain injuries, had lost the emotional signals that normally accompany decisions. Their logic was intact. Their lives fell apart anyway, because they could no longer feel which option mattered. The body, it turns out, votes first, and reason often just ratifies the result.

The shorthand he gave us is the somatic marker — the flicker of bodily sensation that tags a situation as good or bad before the conscious mind has caught up. I should be honest that the details of his theory are still argued over, the experiments have been challenged and refined, as good experiments are. But the core has held, and you already know it is true, because you have felt the decision in your gut before you could justify it on paper.

Now put that next to a finding from psychology that deserves to be far more famous than it is. When researchers put people in a scanner and showed them distressing images, the fear centers of the brain lit up, as you would expect. But when they asked people to put the feeling into words, to simply name it, that activity dropped. Naming a state changes the state. The act of finding the word does measurable work in the nervous system. It is not decoration on top of the experience. It is regulation of the experience.

This is the whole argument of this essay, compressed into a single physiological fact. The discomfort so many of us feel around AI — the one that has no name — is the somatic marker firing in a situation our language has not caught up to. We feel the paradigm shift in the body before we can speak it. And until we can speak it, we cannot quite settle, because the naming is part of how a nervous system metabolizes a change.

Feynman's test

Richard Feynman, the physicist, had a habit that has stayed with me as a kind of conscience. He distrusted the difference between knowing the name of something and knowing the thing. You can learn that a bird is called a warbler in five languages, he liked to say, and know nothing whatever about the bird. The name is not the knowledge. The name can even be a way of avoiding the knowledge, a place the mind stops because it feels like it has arrived.

I think we are at risk of doing this with AI at the largest possible scale. We have the name. Artificial intelligence. It is on every front page and in every strategy deck, and the very confidence of the phrase invites us to stop, to feel we have understood the thing because we can label it. McKinsey reported this year that 88 percent of organizations now use AI regularly in at least one function. And yet only about a third have moved past pilots into anything you could call real transformation. Almost everyone has the tool. Almost no one has the understanding. The name has run ahead of the knowing.

Feynman's test is brutal and simple. Can you explain it without the borrowed word? Can you say what is actually happening, in plain language, from the ground up? When I try to do that with my own late-night unease, the sentence that comes out is not I am using an AI tool. It is something closer to this. I am watching a machine perform something I used to think required a person, and I do not yet know what that makes me.

That sentence is uncomfortable. It is also, finally, true. And the discomfort of it is different from the nameless kind. It is the discomfort of a thing seen clearly, which is the only discomfort you can actually do something with.

The work of finding words

I do not have a tidy resolution, and I distrust essays that pretend to. What I have is a conviction that has only hardened since that night at the keyboard.

The most important skill in a paradigm shift is not learning the tool. The tools will change again before you have finished mastering them. The important skill is the older, slower, stranger one of finding accurate words for what is happening to you, especially when the available words are wrong. Not jargon. Not the borrowed confidence of artificial intelligence and disruption and transformation. The actual words for the actual experience, including the parts that feel like loss, including the unease that does not resolve into a slogan.

This is what I meant, on the front page of this site, by building vocabulary for AI transformation. I did not mean a glossary. I meant the harder thing. The practice of staying with a feeling long enough to name it honestly, so that the nervous system can settle and the mind can finally get to work. Wittgenstein was right twice. The limits of our language are the limits of our world. And those limits are not a wall but a frontier, made and remade by the people willing to do the work of speaking.

The machine, whatever we end up calling it, cannot do that part for us. The finding of the word is the one piece of the building that has to stay ours.

This essay follows from The skills of Captain Future, on building something yourself when the building is shared. If you want to go further than reading, the Knowledge Lounge has a learning path on the same questions.