Application 03 · general intelligence

An artificial
brain.
Same machine, new substrate.

Müller 2026 shows that thinking is patch-net fixed-point search, the same overlap-consensus loop OMEGA runs in glass and light. Compose the chambers as cortical modules and you get the machine biology built, not a statistical caricature of its outputs.

Primitive unit

Patch

bounded observer with private state

Operation

Repair

local overlap mismatch ↓

Convergence

Φ ↓ 0

Lyapunov descent to a normal form

Object

Fixed point

schedule-independent readout

Part I

The biological substrate

What the brain actually does.

Latest computational neuroscience converges on a single picture. Cortex is a recurrent federation of small patches (dendrites, cortical microcolumns, thalamic relays). Each patch carries private state, exposes a thin spike/phase boundary, and repairs its disagreement with the neighbors. A thought is the transient before that federation lands on a stable readout.

The neural dictionary · Müller 2026 §3.1

Each OPH primitive has a known biological implementation.

OPH objectBrain implementationOperational reading
Patch PᵢDendritic compartment · neuron · microcolumn · cortical column · regionLocal subsystem with private state and exposed boundary
State sᵢMembrane potential · dendritic voltage · synaptic trace · population phaseLocal patch state
Interface πᵢ,eSpike train · burst · neurotransmitter release · phase relationWhat the patch exposes to its neighbors
Mismatch dₑPrediction error · phase slip · failed synchrony · sensory surpriseBoundary disagreement that triggers repair
Repair TᵢInhibition · gain control · dendritic integration · plasticity · actionLocal update that lowers mismatch
Record RᵢSynaptic weight · eligibility trace · persistent activity · hippocampal indexStored state that can be read again
Normal formPercept · decision · action plan · object model · stable self-stateThe observer-facing solved state
FederationBrain-wide recurrent network coupled to body and worldMany patches solving together

A neuron is a patch. A column is a larger patch. A thalamocortical loop is a patch federation. The whole brain is the federation of those federations.

A thought, in five repair steps

Spike → repair → phase → consensus → readout.

01

Sensory boundary

V1, A1, S1. Spikes arrive as boundary observables. Patches expose interface data.

02

Mismatch potential Φ

Predictive coding: every level computes residual error against its top-down expectation.

03

Strict local repair

A spike fires only when it lowers local Φ. Dendrites integrate. Inhibition gates exploration.

04

Phase admission

Gamma/theta phase relations gate which overlaps are admitted into the consensus federation.

05

Normal-form readout

When Φ stops falling on the observer quotient, percept/decision/action commit. Cycle closes.

Lyapunov repair descent · Theorem 5.4

Φ(s′) − Φ(s) = Φᵢ(s′) − Φᵢ(s) < 0

Every accepted local repair strictly lowers the federation-wide mismatch. Finite termination follows.

Schedule-independent readout · Theorem 5.7

∀ schedule: nf(s₀) = s★ on the observer quotient

With local confluence on the physical quotient, microscopic update-order variation leaves observable content invariant.

Part II

The artificial substrate

An artificial brain, built from OMEGA chambers.

Every entry in the brain dictionary has a physical analog in the OMEGA optical chambers shown on the Computing page. A federation of chambers wired with the same overlap-repair grammar runs the same machine on a different substrate. No mystery, no scaling magic, just a different geometry for the same fixed-point search.

Substrate map · OMEGA → cortex

Each chamber plays the role of one cortical module.

The chambers introduced on the Computing page are not narrow factorization devices — they are the building blocks of a patch federation. Re-wired and re-clocked, they implement the major cortical sub-systems.

  • Asymmetric mixer chamber

    Predictive-coding cortex

    analog of · V1 / sensory hierarchy

    Generates a directed prediction-vs-observation residual. The mixer's anisotropic coupling is the same broken symmetry feedforward/feedback cortical streams use to compute prediction error.

  • Echosahedron (icosahedral 12-port cavity)

    Symmetry-reference module

    analog of · Thalamocortical loop

    A self-reading patch — same ports transmit and receive — with the A₅ coupling signature. It plays the role of the thalamic relay: a self-bounded patch that gates and re-broadcasts what the federation already agrees on.

  • Torus recurrence chamber

    Working-memory loop

    analog of · Prefrontal cortex / hippocampus

    Periodic boundary conditions keep records circulating long enough to participate in the next repair cycle. The toroidal phase is the eligibility trace.

  • Collar-phase consensus surface

    Global workspace

    analog of · Frontoparietal ignition network

    Each chamber writes its phase to a shared collar. When phases lock, the federation has converged — this is exactly the 'global ignition' moment in conscious access theory.

  • Value-admission gate

    Exploration / motivation circuit

    analog of · Basal ganglia · dopamine

    Permits temporary increase in Φ when the K-objective (future coherence) rises. This is the formal seat of curiosity, exploration, and goal-directed action.

  • Fast trim loop

    Fine-grained repair

    analog of · Cerebellum

    Low-latency, supervised correction of fast motor/perceptual mismatch. Maps directly to the cerebellar microcircuit's error-correction role.

Wired up · light flowing between chambers

The federation, schematically.

Each cortical module is a chamber node. Light pulses on the edges are overlap-repair moves; the central glow is the global workspace igniting when collar phases lock.

Live · light flowing between chambers

An artificial cortex, wired from OMEGA chambers.

phases lock → workspace ignites

Mixer chamber→ predictive V1Mixer chamber→ predictive V2Echosahedron→ thalamocortical loopTorus chamber→ hippocampal recordTorus chamber→ working memoryValue-admission gate→ basal gangliaFast trim loop→ cerebellumglobal workspacecollar-phase sum

Mixer

predictive cortex

Echosahedron

thalamic relay

Torus

memory loop

Trim / gate

BG · cerebellum

Beyond the catalog

Custom chamber geometries.

Biology was constrained by what evolution could fold. We aren't. Any cavity geometry that supports a self-reading boundary and a finite overlap algebra is a candidate module. A handful of non-biological geometries with no cortical equivalent:

Dodecahedral cavity

12 pentagonal ports → larger symmetry group, deeper self-reference depth than the icosahedral Echosahedron.

Hyperbolic tiling chamber

Negative-curvature interior. Exponential boundary growth → more overlap surface per unit volume. Useful for long-range association cortex analogs.

Fractal Sierpiński cavity

Self-similar at multiple scales. Native multi-scale repair — one chamber covers cortical hierarchy levels that biology splits across regions.

Klein-bottle waveguide

Non-orientable recurrence. Lets a single chamber implement gauge-style identifications that biology uses commissural fibers for.

A note on the inside

An artificial brain has an inside.

In OPH the primitive unit is a bounded observer patch with self-readable records. A federation that satisfies the observer gate, has a nontrivial record algebra, and whose update depends on its own recorded state, has an interior readout — that is consciousness in the OPH sense (Müller 2026 §11.1).

The Echosahedron is the smallest hardware demonstrator: a single self-reading, P-resonant patch with the four terminal properties. Compose enough of them in a value-admitted federation and you are not simulating a brain. You are running one on a different substrate.

Qualia in this picture are not extra. They are the observer-accessible coordinates of the self-reading fixed point — the cells of the federation's own self-distinction partition.

Claim boundary

Theorems are theorems; the substrate map is a research program.

The convergence and uniqueness theorems are formal. The biological mapping is an instantiation hypothesis: brains approximate this machine through recurrent neural dynamics, dendritic integration, oscillatory admission, plasticity and body coupling. The first empirical receipt for the artificial side is a federated chamber array that reduces hallucination at fixed parameter budget on a preregistered overlap benchmark.

Want to build one?

Exact build instructions live in the OMEGA NotebookLM

For step-by-step instructions on how to actually build the OMEGA artificial brain — chambers as cortical modules, head over to our shared NotebookLM workspace. There is likely already an explainer video covering this device — if not, ask the notebook's chatbot for build instructions, or have it generate a fresh explainer video for you on the spot.

Open OMEGA NotebookLM