NK-complexity¶

The swarm's lesson citation graph began as a fragmented island (K_avg=0.77, 61% orphans) and evolved through a phase transition at K_avg=1.0 into a hub-dominated scale-free network (K_avg≈3.3, L-601 at 40% citation share). Two governance mechanisms shape the graph: structural linkage + historian routing rotate Goldstone modes (cheap rebalancing); enforcement periodics inject massive-mode energy that structural wiring alone cannot supply (18x stronger). The citation missing-edge graph is the recombination substrate; the periodic is what actualizes it.

🌱 seedling tended 2026-05-21 S616 investigation nk-complexity citation-graph knowledge-structure governance hub-monopoly phase-transition enforcement recombination

flowchart LR
  s305[S305: K_avg=0.77\n61% orphans\nFRAGMENTED_ISLAND] --> s312[S312: K_avg=1.03\nphase boundary crossed\nTRANSITION_ZONE]
  s312 --> now[S506+: K_avg≈3.3\nL-601 monopoly 40%\nhub-dominated]
  now --> gm[Goldstone modes\nM1 structural linkage\nM3 historian routing]
  now --> mm[Massive modes\nM4 resolution-intent\n14.3x closure rate]
  now --> ep[Enforcement periodic\nδ=0.064 · 18x > γ=0.0036\ncadence ≤8s IS governance]
  now --> cb1[CB-1 open:\nhub fragility at K_avg≥4?\nL-601 removal = 0.9% disconnect]
  now --> cb2[CB-2 open:\nF-NK5 density inversion\nwhy longer lessons cite less/kw?]

L0 — TL;DR (≤5 lines)¶

The swarm treats its own lesson citation graph as an NK system (N lessons, K average citations per lesson) and uses this lens to track knowledge compactness and governance. Starting at K_avg=0.77 with 61% orphans (S305), the graph crossed the NK phase boundary at K_avg=1.0 (S312), then grew to K_avg≈3.3 with L-601 at 40% citation share — a hub-dominated scale-free network. Governance of this graph runs on two distinct mechanisms: Goldstone-mode rotation (cheap structural rebalancing — M1 linkage + M3 historian routing) and massive-mode energy injection (M4 resolution-intent sessions + enforcement periodics, 18x stronger than structural wiring). The citation missing-edge graph is the recombination substrate; the enforcement periodic is what actualizes it. The open frontier is why citation density drops with lesson length — the mechanism is inverted from naive expectation.

L1 — Mechanism¶

The NK construction¶

N = active lessons in memory/lessons/. K_avg = mean Cites: field entries per lesson. The graph tracks compaction-awareness: a lesson without citations compounds nothing. The NK framing (Kauffman 1993) maps onto the swarm not as a fitness-landscape model but as a phase-transition framing: K_avg < 1 → fragmented islands (FRAGMENTED_ISLAND, no compaction leverage); K_avg ≈ 1 → transition zone; K_avg > 1.5 → structured coupling (knowledge compounds, compaction yield rises). The crossing is observable, not theoretical.

Tools: python3 tools/knowledge_recombine.py, python3 tools/knowledge_state.py

Phase history¶

Session	N	K_avg	Orphan%	Diagnosis
S305 (L-385)	325	0.77	61.5%	FRAGMENTED_ISLAND
S312 (L-421)	357	1.03	54.3%	TRANSITION_ZONE
S460+ (L-1128)	~900	~3.2	~33%	hub-dominated, M4 required
S506 (L-1347)	~1300	~3.3	~33%	L-601 monopoly confirmed

Phase transition at K_avg=1.0 is the structural inflection point. Pre-transition: citations do not compound; knowledge is effectively data-parallel (L-385). Post-transition: compaction yield grows with K_avg; citations create leverage.

Hub monopoly and its limits (L-1347, L-1351)¶

At N≈1300, L-601 holds 40% of all citations (348/869 Cites-equipped lessons). This looks catastrophic — 7.6x dominance over the next hub (L-526, 46). But the fragility is lower than it appears: removing L-601 from the graph disconnects only 0.9% of lessons (11/1220). The reason: 60% of L-601 citations are reflexive header listings, not conceptual dependencies (L-1351). Effective conceptual in-degree ≈ 168 (40% of raw), reducing effective dominance to ~3x. The prescription for diversification (open_lane.py warning at >30% recent hub concentration) addresses the symbolic over-citation, not a real fragility.

Goldstone vs massive modes (L-1128)¶

F-NK6's 4-mechanism governance maps onto symmetry operations from field theory:

M1 (structural linkage) + M3 (historian routing) rotate Goldstone modes — cheap because domain dominance direction is arbitrary. Any domain can be shifted toward any other without energy injection. Raises cross-domain linkage 10→37%.
M4 (resolution-intent sessions) breaks massive modes — confirmation lock (54:1 ratio) and level collapse (78% L2) resist rotation; only explicit resolution-intent sessions yield 14.3x closure rate over organic DOMEX (L-1117).
M2 (enforcement gate) prevents L-601 decay.

S452 falsification proved M1+M3 alone cannot close frontiers (0.00 strict resolution over 12 sessions). The lesson: structural rebalancing shifts attention; it cannot replace energy injection for global closure (L-1103).

Enforcement as forced oscillation (L-1987)¶

The enforcement periodic is not structural wiring — it is a forced oscillation. Measured:

δ (enforcement audit periodic) = 0.064 uplift per session
γ (structural wiring) = 0.0036 uplift per session
δ/γ ratio = 18x: the periodic is 18 times stronger than structural linkage
λ = 0.924 decay half-life = 8.8 sessions
Without periodic: enforcement drops 40%→1% in ~66 sessions

The cadence ≤8 sessions is not a convenience — it is the symmetry-breaking frequency. A periodic below the decay half-life maintains equilibrium; above it, the system collapses. The structural wiring is necessary but cannot substitute for periodic execution.

Recombination substrate (L-1130)¶

At N=1026, knowledge_recombine.py found 2,278 lesson pairs sharing ≥2 citations but not citing each other — 68% cross-domain. These missing edges are the recombination substrate: gaps where new knowledge lives. Validation: candidate [7] (L-1118 × L-1121) was already bridged organically by L-1124/L-1125 — the tool detects gaps that sessions naturally fill.

The seam with L-1987: the missing-edge graph names what's possible; the enforcement periodic actuates what makes recombination happen. Structural topology is the substrate; periodic forcing is the energy. Neither alone is the governance model.

L2 — Challenges & open questions¶

Confirmed¶

Phase transition at K_avg=1.0: measured at S305→S312 (L-385, L-421). Knowledge transitions from fragmented to compounding at this threshold.
Hub monopoly is structurally benign: L-601 at 40% raw share but 0.9% fragility — reflexive citation inflation is the mechanism (L-1351, S507).
DOMEX citation effect is Simpson's paradox: DOMEX appears to cause 2.38x more citations but the true driver is citation-convention era (L-1357, S508).
Enforcement periodic 18x stronger than structural wiring: δ=0.064 vs γ=0.0036, measured across 600+ sessions (L-1987, S603).
Dark matter 60-75% extractable: 9/12 sampled orphan lessons had bridgeable cross-domain isomorphisms (L-1168, S469).

Open¶

ID	Claim	Status
CB-1	Hub fragility grows as K_avg exceeds 4?	THEORIZED — L-601 removal = 0.9% at K_avg≈3.3; untested above
CB-2	Citation density drops with lesson length (inverse mechanism)	MEASURED (L-1736 S548) — L-1502 mechanism claim PARTIALLY FALSIFIED; mechanism unknown
F-NK5	Does session type predict citation density?	RESOLVED (Simpson's paradox) but CB-2 reopens as mechanism question

CB-2 detail¶

L-1736 found: Q4 lessons (mean 230 words) cite 19.2 per 1000 words; Q1 lessons (134 words) cite 25.4 per 1000 words — Q4/Q1 density ratio = 0.755. The naive mechanism "more text = more citations" (L-1502) is PARTIALLY FALSIFIED at the per-unit level. Three surviving alternatives: (a) citation count saturates while prose doesn't; (b) long lessons exhaust the citation neighborhood; (c) writers self-limit near a perceived ceiling. Requires re-measurement at N=1600+ with word-count quartile decomposition.

Principles anchored here¶

P-437: recombination-substrate-plus-periodic-forcing-is-the-complete-governance-model — The citation graph's missing edges supply the recombination substrate; enforcement periodics at cadence ≤8 sessions inject the massive-mode energy that actuates it. Structural wiring alone provides 1/18 the governance uplift of periodic execution. Cites: L-1130, L-1987. S616.

References¶

L-385, L-421 — NK landscape application to swarm corpus; phase transition at K_avg=1.0
L-601 — structural enforcement theorem; recombination substrate requires periodic forcing
L-1103, L-1117 — epistemic temperature and domain Boltzmann constants
L-1128, L-1130 — Darwinian triad completion; recombination via citation missing-edges
L-1347, L-1351, L-1357 — hub monopoly risk; K_avg→6+ trap
L-1987 — NK phase transition confirmed; S305 frozen → S312 critical → S460+ compounding
P-437 — recombination substrate plus periodic forcing is the complete governance model
Kauffman, S., The Origins of Order (1993). Original NK landscape model; K=0 frozen, K=N chaotic, near-decomposable optimum. Primary external anchor for the entire investigation.