Gap 2 — cortical microstructure as a hub topology

Gap 2 gives the cortex the internal microstructure Sim 1 left as one node. The fill is the hippocampus precedent re-organized into a hub topology: seven long-range hubs carry all long-range edges. Cutting the hubs collapses distant binding while a mass-matched local cut spares it; ST-2 confirms the phenotype is topological and sign-stable across nine cells.

The additive cortical micro-model fills Gap 2 by applying the hippocampus precedent to cortex: a frozen autoassociator sits behind one node, but its connectivity is re-organized into a hub topology — N_HUBS = K = 7 (the emerged working-memory capacity) long-range hubs carry all long-range edges, non-hubs wire only locally, and the hub-mediated recurrent settle is the reentrant loop. Cutting the hub edges collapses distant binding (0.857 → 0.724) while a mass-matched random-local cut spares it (→ 0.850); the effect vanishes all-to-all (topological) and is long-range-specific. ST-2 is sign-stable across 9 cells, so the micro-model is sufficient and the kernel fork is not triggered.

The gap: one undifferentiated node

Sim 1 carried the entire cortex as a single node of the 12-organ kernel — Hard Limit 1. Gap 2 gives that node the internal microstructure executive function needs, and it does so the additive way, so the frozen engine is untouched and the M9 anchor stays bit-identical.

The fill: the hippocampus precedent, re-wired into hubs

The micro-model is the hippocampus precedent applied to cortex. A frozen autoassociator sits behind one node, but its connectivity is re-organized into a hub topology: N_HUBS = K = 7 (the emerged working-memory capacity) long-range hubs carry all the long-range edges, non-hubs wire only locally within radius r = N/(2K) = 9, and the hub-mediated recurrent settle is the reentrant loop. Everything is grounded (FOXG1 γ = 1.4737, emerged K = 7, barrier(FOXG1) = 0.542948); new_tuned_constants = 0; the engine is read-only.

The read-outs are non-circular by construction. The distant-binding read-out cues a near block and measures recovery of a distant block — never the perturbed edge — and a mass-matched random-local cut of equal mass is the control.

What the build found

Cutting the hub long-range edges collapses distant binding (0.857 → 0.724) while a mass-matched random-local cut spares it (→ 0.850); the hub−random gap is monotone and sign-stable (0 → +0.126). The same cut on an all-to-all substrate gives a gap of ≈ 0 (0.0019) — the effect is topological, not generic mass. A scattered (non-local) cue gives only a weak gap (0.009) — the binding is long-range-specific.

The honest residual is that the cell-gain gene never enters the dynamics: LHX2 == FOXG1 to the read-out, because g does not appear in the sign() update, so the phenotype is purely topological / gene-blind. Absolute magnitudes and gene-specificity stay [O].

ST-2: non-circularity and set-shifting

The break ST-2 was designed to cause is that the frontal phenotype only appears in a circular read-out, or that set-shifting flips under a duration sweep. Had it broken, the additive micro-model would be insufficient, and the major fork — resolving cortex inside the kernel, which breaks M9 — would open. The test imports the exact build model (non-circular at the code level) and tests three signed features across 9 cells (3 seed cohorts × durations 20/40/60): a non-circular behavioural bind-gap (shown against the circular hub-mass-drop control: hub-cut drops 0.60 vs random-cut 0.000), a topological gap (bind-gap exceeds the all-to-all control), and a hub-lesion perseveration delta (establish attractor A, cue B on the near block, and measure the distant block still resting on A).

What survived, and what stays open

The phenotype survived: phenotype_sign_stable = True, broke_on = [], perseveration delta +0.033 → +0.172. Gap 2 is graded [L] in-silico; the additive micro-model is sufficient, the kernel fork is not triggered, and M9 stays bit-identical. The set-shift delta is duration-invariant because the autoassociator settles fast, so the sweep confirms no flip (anti-tuning) rather than duration-dependence; the gene-blindness is carried forward as the standing [O].

The additive discipline held — re-verified at this session, the M9 anchor reproduces bit-for-bit, the engine tree is unchanged, and the M0–M16 subtree is unchanged.

Firewall (Axis-A). Every quantity on this page is a structural quantity of an in-silico model — a sign, a margin, an ordering — and is not a felt state, an experienced quality, or a level of consciousness (consciousness_claim = 0; the hard problem stays open, hard_problem_open = 1). These are model results, not validated neuroscience and not clinical guidance. efficacy = 0; not medical advice.