QESIS+ v8.6 (2026-08-09) · method, not marketing
This page carries the qualitative method behind the index. Choose the causal conditions, move the calibration anchors, and the configurational analysis recomputes in your browser over all 35 states. If your assumptions produce a different answer, that is the point.
Regression asks which variable moves the outcome on average, holding the others still. That question assumes the causes are independent and that more of one thing always means more of the outcome. Substrate sovereignty is neither. A state can be water rich and cable poor, or cable rich and captured at the platform layer, and the combination is what decides the outcome rather than any single axis.
Fuzzy-set Qualitative Comparative Analysis asks a different question: which combinations of conditions are consistently sufficient for the outcome. It treats causation as conjunctural, asymmetric and equifinal, meaning several distinct routes can reach the same result and the route to the outcome is not the mirror of the route away from it. That matches the object. It is also falsifiable in a way an R-squared is not: a single case inside a sufficient configuration that fails to show the outcome damages the claim, and you can go and look for it.
Pick conditions on the left. The truth table, the solution and the trilemma reading recompute as you move. Calibration is the analysis, so the anchors are exposed rather than buried: change them and watch the solution move.
Two to four is the usable range. Five conditions is 32 rows, six is 64, and beyond that the table has more configurations than the world has states to fill them.
Every row is shown, including the ones that fail. A table that hides its failures cannot be audited.
| Configuration | Cases | Consistency | PRI | Coverage | Verdict |
|---|
Computed live from QESIS+ v8.6 (2026-08-09) axis scores. Consistency, coverage and PRI follow Ragin. Minimisation is conservative: no counterfactuals on logical remainders, so what you see is the safe solution rather than the tidy one.
A condition is conventionally called necessary at consistency 0.90 with coverage 0.60. Anything weaker is reported and not called necessary.
| Condition | Consistency | Coverage | Reading |
|---|
Each solution is read against three constraint structures. Two are established in political economy. The third is the contribution this work makes, and the reading is strongest where all three point the same way.
Deep economic integration, the nation state, and mass democratic politics. A state can hold two.
Rodrik, D. (2011), The Globalization ParadoxA fixed exchange rate, free capital movement, and an independent monetary policy. A state can hold two.
Mundell (1963) and Fleming (1962)Algorithmic independence, hyperscale efficiency, and ecological sustainability. A state can hold two.
Batista Silva, R. (2026), Ontological Blind Spots. Sovereignty measured as metabolic capacity, not legal status.A configuration is not only a statistical object. It describes a position a state occupies against others, and positions have payoffs. Read as a game, the substrate layer has a structure worth naming.
Hyperscale capacity is a coordination good with strong increasing returns, so the dominant strategy for any single state is to attach to an existing hyperscale network rather than build a sovereign one. That choice is individually rational and collectively produces exactly the concentration the index measures. It is a stag hunt rather than a prisoner's dilemma: cooperation on a sovereign alternative pays better for everyone than universal defection to a foreign platform, but only if enough states move together, and no state moves first alone. The European sovereign cloud efforts are that coordination problem played out in public.
The physical axes behave differently. Water and electricity are not coordination goods, they are capacity constraints, and no amount of coordination relaxes them. That asymmetry is why the composite weights the physical axes heavily: a state can renegotiate its platform dependency, and cannot renegotiate its aquifer.
Each axis names the institutions that produce the underlying measurement, with the dataset and its metadata as separate links. Reading a number and checking a method are different tasks and deserve different doors.
| Axis | Trilemma vertex | Institutional sources |
|---|---|---|
| WSE Water Stress Exposure |
ecological sustainability |
|
| CSE Cable Stress Exposure |
algorithmic independence |
|
| REE Rare Earth Element Stress |
hyperscale efficiency |
|
| FPE Foreign Platform Exposure |
algorithmic independence |
|
| ODI Operator Dependency Index |
algorithmic independence |
|
| CRD Cloud Risk Density |
hyperscale efficiency |
|
| ESE Electricity Stress Exposure |
ecological sustainability |
|
Those three words are cheap. Here is what each one means concretely, and how to check it.
Every agent operation is risk scored before
it runs, hashed on the way out, and appended to a chain where each entry commits to its
predecessor. Database triggers refuse deletion and refuse any edit except recording a human
approval. A retroactive change breaks the chain downstream where an auditor sees it.
EU AI Act Art. 12 ISO 42001 A.8.2
Every asset carries an identifier, a SHA-256
and a lineage note, and the register is re-checkable end to end: each entry recomputes from
the file it names. Material whose licence was never recorded is reported as unrecorded rather
than tidied away, and cannot reach a public artifact.
EU AI Act Art. 10 ISO 42001 A.6.1
The engine on this page is the same algorithm
that produced the published solution, ported to run in your browser with no server. Anchors
and cutoffs are exposed because calibration is the analysis. Same inputs, same output, every
time.
EU AI Act Art. 11 Annex IV
The EU AI Act sets the obligations this system is built against: Article 9 requires risk management across the lifecycle, so risk is scored before an operation runs rather than reviewed after. Article 10 requires data governance, which is the asset register and the licence discipline. Article 11 and Annex IV require technical documentation, which is generated from the live registry rather than typed, so it cannot go stale. Article 12 requires automatic logging and traceability, which is the hash chain. Article 14 requires human oversight with a real ability to intervene and stop, which is the gate that currently holds every high-risk output. ISO/IEC 42001 wraps those obligations in a management system: A.6.1 data quality and provenance, A.6.2 risk assessment, A.8.2 traceability of decisions, A.9.3 human oversight and intervention. Neither framework is treated here as a post-deployment audit. Both are architectural constraints, which is why the compliance tables exist in the schema rather than in a folder.
Stratify the sample by jurisdiction tier and rerun. If the cross-axis associations collapse inside every tier, the pooled coefficients were composition rather than mechanism, and the composite would need rebuilding on tier-relative terms. Separately, a single state sitting inside a sufficient configuration while showing low outcome membership damages that term directly. Both tests are runnable on this page: narrow the conditions, move the anchors, and see whether the solution survives your assumptions rather than mine.