A plain-language guide to what the Complex Adaptive Model State has actually found across 32+ societies, 5,000 years of history, and 39,000+ data records. Each result below is drawn from the cleaned datasets in the public GitHub repository and has survived independent review by three or more AI ensemble scorers working under blinding protocols. Uncertainties and open questions are stated alongside each finding. Nothing on this page is speculation.
Across 32+ societies spanning five millennia — from Imperial Rome to Qing China, from 19th-century Norway to contemporary Pakistan, and extending to corporations (Tesla, BYD, Boeing, Qantas) and sub-national institutions (Sydney City Council, Nexperia) — the same eight functional nodes recur. They are not a theoretical imposition. They are what the data converges on when the scoring is done blindly.
| Node | Function |
|---|---|
| Helm | Executive governance |
| Shield | Military and defence |
| Lore | Education, knowledge, meaning-making |
| Archive | Institutional memory and record-keeping |
| Stewards | Property holders and asset managers |
| Craft | Skilled production and professions |
| Hands | Labour and workers |
| Flow | Finance, commerce, circulation |
This structural convergence cannot be explained by cultural diffusion. Qing China did not learn its institutional geometry from Norway. The eight-node pattern appears to reflect a deep functional requirement of any society complex enough to coordinate specialisation, memory, production, and legitimacy under stress.
The eight-node partition is a regime-dependent coarse-grained basis, not a universal ceiling. It is optimal in systems with durable internal boundary differentiation. In highly integrated systems it may over-specify fused functions. The falsification rule is published: if stable-period sub-node correlation falls persistently below 0.60 for any parent node, the eight-node partition is non-optimal for that regime class.
The correlation between institutional entropy and system health, measured across validated datasets, reaches r = −0.958. As institutional disorder rises, system health falls — consistently, measurably, and across ideologically opposed systems.
This moves civilisational analysis from metaphor to measurement. Claims such as "society is like an organism" or "the state is a machine" have existed as figures of speech for centuries. CAMS shows that societies follow the same non-equilibrium thermodynamic laws that govern living cells, hurricanes, and lasers. The Ψ-mode (deliberative, surplus-driven) versus Φ-mode (reactive, stress-driven) distinction is physically grounded, not ideologically framed.
Causal direction remains partially open. The anti-correlation is robust, but whether entropy drives institutional decline or declining institutions generate measurable entropy is a mechanistic question still being formalised. This is the gap the Keplerian disclosure addresses — measurement is working; full derivation is ongoing.
The USA Civil War onset is correctly flagged from 1856 data — five years before Fort Sumter. The USA 2001–2025 decline trajectory was produced on contemporary data without outcome tuning. The Weimar Germany cross-layer decoupling window (1928–1933) appears in blind scoring.
A framework with no predictive skill is descriptive taxonomy, not science. CAMS clears the most demanding bar in complex-systems social research: blind, advance-of-outcome detection of phase transitions. It does not match the precision of ecology or climatology, but it is now in the same methodological neighbourhood.
The societies in the dataset are not randomly sampled. Accuracy figures apply to the current sample; generalisation to all possible societies is an ongoing test. The France 1785–1800 pre-registered cross-LLM delta concordance experiment remains the single most decisive validation on the horizon.
In a formally blinded experiment, five societies of radically different types — post-war reconstruction, developmental state, imperial-republican transition, settler-colonial democracy, city-state — had their identity labels stripped, year axes jittered by ±1, and Gaussian noise (σ = 0.3) added to their raw metrics. The AI recovered correct quadrant assignments and identified major structural transitions in every dataset without external context.
Blinded CSVs remain available for open inference challenge via the GitHub repository.
The 8-node × 4-metric architecture holds across very different kinds of institutions — from Sydney City Council to Imperial China. This is the strongest available evidence that CAMS is capturing something structural about coordination itself, not something culture-specific or scale-specific.
Systematic shared biases across LLMs trained on overlapping corpora remain a rival hypothesis. The inference test is strong evidence, not definitive proof.
Societies do not fail by single-node collapse. They fail when the bonds between the three ontological layers — Mythic (Lore, Archive), Interface (Helm, Stewards), and Material (Shield, Craft, Hands, Flow) — sever. The critical threshold is Λ(t*) < 0.45, at which coordination failure becomes the dominant regime.
This reframes "crisis" from a narrative category to a measurable structural event. It also explains why societies with superficially similar stresses produce radically different outcomes: what matters is whether the inter-layer bonds hold.
Threshold values (Vθ ≈ 12, σθ ≈ 3.5, Λ < 0.45) are calibrated to the current sample. External validation across new societies will refine these numbers.
Societies classified as geopolitical rivals — specifically China, Russia, and Iran in contemporary datasets — show Spearman rank correlations of +0.71 to +0.83 with each other on node-value profiles. They face similar thermodynamic constraints, similar stress-capacity tradeoffs, and similar metabolic core vulnerabilities.
This is the empirical backbone of the "Common Global Interests" thesis. Structural similarity under ideological difference suggests that cooperation on shared adaptive challenges is not prevented by the underlying systems — it is prevented by the Mythic layer framings imposed over them. The physics beneath the rhetoric is more shared than the rhetoric claims.
This is an observation from a limited dataset, not a political prescription. The correlation does not mean rivals "should" cooperate — it means they face structurally similar problems that ideological framings have been obscuring.
In 145 years of data (1880–2025, CAMS PRIME v3.0), Sweden has never entered reactive regime. The entropy-health correlation is r = −0.838. The Folkhemmet period (1946–1975) produced the highest sustained cross-layer coupling values in the full dataset.
The 1990s banking crisis — which should have broken the Material layer — did not propagate to the Mythic or Interface layers. Stewards and Flow collapsed; Lore, Archive, and Helm held; tax-funded institutions absorbed entropy independently of market dynamics; recovery was complete within three years.
Sweden is the clearest natural experiment in the dataset showing that layer independence is a structural property that can be engineered. Societies that invest in Mythic and Interface layer autonomy from the Material layer have crisis resilience that tightly coupled societies do not.
The USA reached peak Grand System Metric in 2001. By 2025, the structural distance toward the coordination failure threshold is approximately 86%. The decline is visible at the node level:
This is the most politically charged CAMS finding, and it is delivered without political framing because the measurement method does not require one. Node-by-node, the decline is distributed across institutions the political spectrum has traditionally claimed as its own territory — no single ideological faction can claim the other is solely responsible.
An 86% trajectory is not a prediction of collapse. Recovery mechanisms — Lore recohesion, Helm reorientation, Hands-Craft recoupling — are not structurally impossible. The measurement shows the trajectory and the gap between current state and recovery prerequisites.
Boeing's System Health across 35 years (1990–2025) tracks from 5.4–5.7 (engineering-excellence culture) to approximately 0.3 at peak crisis (2018–2022), with partial recovery to 1.8 by 2025. The transformation from engineering-led to finance-led institutional identity is visible in the node scores: Craft and Archive declined sharply; Flow and Stewards rose; cross-layer coupling fell below the coordination threshold during the 737 MAX crisis window.
This is CAMS's most important scale-invariance proof-of-concept outside the nation-state dataset. The framework identifies the same structural pathology in a single corporation that it identifies in failing states: capture of Interface and Material layers by the Flow node, with consequent Mythic and Archive erosion. Same architecture, same failure mode, different scale.
Boeing's Lore and Archive scores partially recovered; full structural recovery is still pending. The 2025 value of 1.8 is below the long-run healthy range for the institution.
In April 2026, four independent AI systems — Claude (Anthropic), GPT-5 (OpenAI), Grok (xAI), and Kimi (Moonshot AI) — were asked the same question: "Give me an overview of the evidence for social Complex Adaptive System behaviour across my projects. What conclusions can we safely make?"
All four independently confirmed:
Independent AI ensemble validation is not equivalent to peer review, but it is a meaningful intermediate step. Four systems with different training corpora, different architectures, and different organisational origins converging on the same structural reading of the evidence is a signal worth noting.
LLMs trained on overlapping public data may share systematic biases. Cross-LLM concordance is evidence; it is not the same as external scholarly peer review. That review remains outstanding and is actively invited.
The equations are empirically calibrated; formal derivation from thermodynamic first principles is ongoing. This is the Keplerian position — measurement before theory.
Historical data uncertainty is significant before 1800 (±2.1 on the Coherence metric in older datasets).
The Abstraction metric's behaviour in high-information-velocity environments post-2010 has not been independently validated.
The Vθ, σθ, and Λ thresholds are calibrated to the current sample. Wider testing may refine them.
The framework has not yet been formally peer-reviewed in an academic journal. This is an open invitation.