Notes toward AGI · Series 01

Toward an AI that
thinks — not one
that predicts.

Three notes on why current systems cannot reason, how biological cognition differs, and the architecture required to close the gap. They form a single argument, cross-linked at every point of contact.

Notes
03
Ideas mapped
09
Shared spine
Structure & process

fig. 00 — the atlas · hover a node

01 — The argument

One claim, carried
across three notes.

Each note answers the one preceding it. Read individually they are observations; read in sequence they constitute a specification.

I.

Prediction is not reasoning

An infant communicates reliably on weeks of experience. A model trained on trillions of tokens cannot derive what its corpus does not already contain.

II.

Topology, not scale

86 billion neurons in a three-dimensional network, under no single governing formula, acquiring capability through continuous interaction rather than optimisation.

III.

Recursion requires structure

Replicating a model and tuning it automates a procedure without making it recursive. Self-improvement requires a system able to operate on its own structure.

03 — The spine

“Neurons, synapses and the remaining physical components constitute the structure — the hardware. Cognition, which we refer to as consciousness, is the process executing on that structure — the software.”
TN-003 · §05, structure and process

One idea recurs in all three notes: the separation between physical structure and the process executing on it. On the atlas above it holds the centre, because every thread passes through it.

04 — Threads

Pull one idea through
all three notes.

Every passage is tagged with the ideas it carries. Selecting one collects it from wherever it appears across the three notes, in reading order.

05 — Correspondence

Corrections are more useful than agreement.

These are working documents rather than conclusions. Specific objections, counter- arguments and related work are all read by the founding team.