01
Language is a brain event before it is a sentence.
Watch the inferior frontal gyrus warm before a word is found. The model lets us see meaning under construction — the half-second when a thought is still gathering itself.
There is a model that predicts what published research says an average brain does.
Scroll to see
How it works · I
TRIBE is a brain-encoding model.
Trained on thousands of hours of fMRI recordings, it learns the map from a stimulus — a sentence, a passage, a piece of music — to the brain's response. Then it predicts.
Eight rooms
We built eight rooms to show you.
Four encoder rooms put the model to work, two depth-psychology rooms drop into older language, two literary rooms read Goethe and Dante alongside the same circuits. Hover a doorway to see the pattern. Step inside when you're ready.
The Instrument
Underneath the rooms.
Twenty regions catalogued one by one, eleven bridges named between them, six guided walks that read short passages with the brain reacting in real time. The source layer for everything above.
The Long Form
Slow reading.
Two ongoing essay series that sit alongside the rooms rather than inside them — where the science gets long enough to argue with itself.
What you'll learn
02
Music moves the same regions that move you.
An ambient drone, a 1923 jazz cut, and a Thai luk thung lullaby all bring different fingerprints to the auditory cortex — and to the default-mode network, the part of you that's still you when you stop trying.
03
What a model can't translate is itself a kind of knowledge.
TRIBE was trained on English. The Thai prompts are not just a different input — they reveal the shape of what the model never learned. The silence is part of the finding.