I ended the last piece with a sentence I want to make good on: the brain builds maps, and not just of physical space.
Start with the physical maps, because those are the ones we can point to in the tissue.
In 1948, the psychologist Edward Tolman made an unfashionable claim. A rat running a maze, he argued, was not chaining stimulus to response. It was building a map. It carried a model of the maze inside it, which is why it could take shortcuts it had never been trained on. The dominant view of the day had no room for an inner representation. Tolman insisted there was one.
It took two generations to find it in the tissue. In 1971, John O’Keefe recorded neurons in a rat’s hippocampus that fired only when the animal stood in one particular place, and went quiet everywhere else. Place cells. In 2005, May-Britt and Edvard Moser found the coordinate grid underneath them: cells in the entorhinal cortex firing in a repeating hexagonal lattice across a space, a kind of biological graph paper. O’Keefe and the Mosers shared the Nobel Prize for it in 2014.
So the brain keeps a literal cartography. That much is settled.
The part that is still being worked out is more interesting, and it is the reason this matters for the inner life. The same machinery does not stay confined to rooms and mazes. When people reason about abstract spaces, relationships between things that have no physical location, the same grid-like signal shows up that appears when a rat crosses a floor. [hypothesis] This is an active frontier, not a closed result, so I hold it loosely. But the direction is consistent: the system that tells you where the door is looks a great deal like the system that tells you how far one idea sits from another.
Now hold that next to the other half of my work, because they rhyme in a way I did not expect the first time I saw it.
In the previous essay I described what a language model actually builds when you train it to predict text. It encodes meaning as position in a high-dimensional space. Words that mean similar things end up near each other. Relationships between concepts become directions you can travel, which is why the analogy “king is to man as queen is to woman” turns out to be literal arithmetic on vectors: take the vector for king, subtract man, add woman, and you land next to queen.
That was not a new intuition, only a newly literal one. The linguist J. R. Firth put it in 1957: “you shall know a word by the company it keeps.” Meaning lives in relationships, in what sits near what. Embeddings took that sentence and turned it into geometry you can measure.
Here is what stops me.
A neural network trained on human text is a few years old. A hippocampus is the product of hundreds of millions of years of evolution. They share no substrate, no origin, no designer, no purpose. And they arrived at the same solution. Both represent meaning as a position in a structured space, and the distance between positions as a measure of how related two things are.
When two systems this different converge on one answer, the convergence is usually telling you something about the problem, not the systems. It suggests that meaning may simply be geometric. To know what something means might be nothing more, and nothing less, than to know where it stands in relation to everything else.
This is the point where the research turns personal, because your inner world is one of these spaces.
If that is right, the ordinary difficulties of a mental life stop being vague and start being spatial, and spatial problems have spatial solutions.
Confusion is often two concepts collapsed into one, sitting too close together when they need to be held apart. Insight is frequently a direction you had not walked yet, a path between two regions that were always there but never traversed. A mood is a region you find yourself occupying. A habit is a path worn deep enough that you slide down it without deciding to. Reframing, the move a good therapist spends years teaching, is at bottom relocation: taking a fact and setting it in a different position, where its neighbors, and so its meaning, are different.
And the tools for moving through the space are not exotic. Attention is how you travel. Repetition is how you reshape the ground. This is the same mechanism I keep returning to in the work on machines: Aristotle’s hexis, the disposition laid down by repeated action until it becomes a structural feature of who you are. The path you walk most often is the one that deepens. A brain, like a character, becomes what it repeatedly does.
So “you can redesign how you think” is not a slogan, and I distrust it as one. It is a fairly dry consequence of meaning being a geometry rather than a fog. You cannot edit a fog. You can edit a map.
That is the premise this whole section rests on, and the premise of the book I am writing. Not that the mind is endlessly malleable, it is not. Not that a better attitude dissolves a hard life, it does not. Something narrower and more usable: the structure is real, it is at least partly yours to shape, and naming it precisely is the first move in shaping anything at all.
The brain spent an age building a machine for finding its way through space. At some point it turned that machine inward and began using it on meaning. The territory is already there, and you are already living in it. The only question left is what you do with it now that you know its shape.