Brains and language on one manifoldA cortex and a language model, reading the same words, trace the same path

The boldest version of substrate-agnosticism crosses the carbon/silicon line. If meaning lives on a universal manifold, then a brain reading a sentence and a language model reading the same sentence should trace the same trajectory geometry through it — not the same voltages or activations, but the same shape of motion: where it turns, how sharply, in what order. (The two ideas doing the work here — rotations preserve shape, dot products measure agreement — are built from nothing in the rotations primer.)

Below, a trajectory derived from EEG (a person reading) and one from a language model (reading the same text), in coherence coordinates. Raw, they sit at different orientations. Press Align (CRT) — the Canonical Resonance Transform, a rotation between substrates — and their tangent dynamics line up: where one turns, the other turns.

Text and EEG trajectories in coherence space. Press Align: the tangent correlation here climbs from ≈ 0.1 to ≈ 0.8 (in the measured EEG↔text data, the curvature correlation reaches ≈ 0.71). Toggle tangent / radial: the semantic signal lives in the tangent (rotational) dynamics — how the trajectory turns — not the radial (how far out it sits). The radial needle barely moves, aligned or not: comparing the wrong component hides the correspondence. That is the “quality inversion.” tangent correlation = —

The measured numbers are a curvature correlation of 0.71 and a temporal correlation of 0.79 between EEG and text trajectories, with the same PAS ≈ 0.76 attractor recurring in cortex. And the diagnostic that the signal is rotational, not radial, is what a tangent-space decoder would exploit to close the gap from today’s ~53% decoding toward the ~79% information ceiling.

Hypothesis-grade — honest statusThis is the least-settled rung of the ladder. The curvature/temporal correlations are real measurements on EEG-while-reading data, but the tangent-decoder program that would turn this from suggestive into decisive is staged, not yet run. We mark it as a frontier: a clean prediction (tangent dynamics carry the semantics) with an experiment designed to falsify it. The synthetic trajectories here illustrate the geometry; the claim awaits that run.

That is the universal case, end to end: one law and one constant from the vacuum to the mind, one shared manifold of meaning across models, and — provisionally — across the carbon/silicon line. All that remains is to stand back and see what it adds up to.