Operator Portability in CAMS — Synthesis & Verification Trail

Can the same eight-node coordination scoring apply to a corporation and a country without cheating? The short answer: the operators transfer cleanly (structural portability), but their calibration does not (parametric non-invariance) — a firm's "normal" bond strength and a nation's "normal" bond strength are different numbers on the same ruler, not the same number.

This page is the entry point to a small cluster of related documents produced together: a synthesis paper drawing the confirmed findings together, the formal Operator Portability Proposition write-up, and two independent computational-recomputation audits that checked every reported statistic against raw data before any of it was allowed to stand. A companion paper on the Soil-Bond Index (η_soil) and its Norway baseline is deliberately not part of the confirmed findings below — the second audit found real problems with it, and it's paused pending a rework. That's reported here too, not hidden.

Headline

Confirmed — canonical CAMS operators (aggregate bond strength B̄, algebraic connectivity λ₂) are structurally portable across a 9-firm corporate panel and two national historical panels (Denmark 1752–2025, France 1850–2026), while remaining parametrically non-invariant: out-of-sample firms run +10.4%/+12.2% higher in B̄/λ₂; France behaves as a "bridge" entity (94.9% containment within the corporate band) while Denmark is a clean "separation" case (19.3% containment). A stricter within-corporate diagnostic further shows that even a single firm-invariant B̄–λ₂ slope is formally rejected (p=0.008) — non-invariance operates within the corporate class, not only across the corporate/national boundary.

1. What's confirmed Confirmed

Two independent recomputation passes — each working directly from raw eight-node Coherence/Capacity/Stress/Abstraction scores, never from a paper's own summary tables — confirm:

Structural portability (P.4.1): B̄–λ₂ correlation ≥ 0.93 across every held-out firm and every entity tested; the computational grammar transfers without needing a different equation per domain.
Parametric non-invariance (P.4.2): Kolmogorov–Smirnov and Mann–Whitney tests reject distributional equality between in-sample and out-of-sample firms, and between each national panel and the pooled corporate distribution.
Threshold relativity (P.4.3): entity- or class-specific percentile baselines are required; a universal absolute cutoff produces uncontrolled false-positive rates. (One status-table entry mis-stated this criterion's own threshold in an earlier draft — corrected in the synthesis.)
Bridge entities exist (P.4.4): France overlaps substantially with the corporate distribution; Denmark does not. Overlap is the exception, not license to pool classes.

2. What's preliminary Preliminary

A corporate functional-variability result — that firms concentrate operational shock-absorption in specific nodes (Craft, for three manufacturing firms; Flow, for a digital-platform firm) rather than suppressing variability outright — is included in the synthesis but explicitly flagged as not yet through the same audit standard applied to the results above.

3. What's paused Paused

The Soil-Bond Index (η_soil) and its Norway baseline are excluded from the confirmed findings. A second, more thorough audit found: reported t-statistics that match neither Student's nor Welch's test on the named files; a Germany source file silently missing its entire 1949–1989 division-era range; a composite metric that discriminates crisis years worse than one of its own numerator components (Archive bond strength alone, AUC 0.907 vs. the composite's 0.861); an internally self-contradicting hypothesis (H2 predicts crises always show high dispersion, but the paper's own "Coherent Collapse" regime is defined by low dispersion); and named "regime attractors" resting on one example society each with no clustering or classification test run. Because the Norway baseline is computed entirely via η_soil, it inherits every one of these open problems. Both are being reworked to the same standard as the confirmed findings above before they're added back.

4. Files & downloads

Open science — the synthesis, the formal proposition write-up, the source manuscript covering the paused η_soil/Norway/functional-variability work, and both audit passes are all available below.

Synthesis paper (Markdown) Operator Portability Proposition (PDF) Norway / functional-variability manuscript (PDF, Norway section paused) Audit v1 (DOCX) Audit v2 — supersedes v1 (DOCX)

Method note. All operators recomputed independently from raw Coherence/Capacity/Stress/Abstraction node scores across two audit passes, never from a paper's own summary tables. Corporate panel: 9 firms, 2006–2026 (4 in-sample, 5 held out). National panels: Denmark 1752–2025, France 1850–2026. The Soil-Bond Index and Norway baseline are excluded from confirmed findings pending rework — see Section 3 above.