v9.1 (2026-09-17) continuously republished 3 states withheld with cause
The instrument that refuses its own bad data.
This endpoint does not sell a dataset. It publishes what a validation pipeline currently certifies, withholds what it cannot verify, and refuses its own release when a published number stops agreeing with the source it came from. The first thing it caught was an error in the study it was built to support.
What the pipeline caught, in its own numbers
An instrument that has never contradicted the person who built it has not been tested. This one has. The QESIS Theory column is not the declared formula: recomputing the declared formula from the axis columns published beside it disagreed with the published column on most rows, and the deviation was sign ordered rather than random. The figure was withdrawn, not softened. Every correction since is carried in the concordance with its evidence, and 2 of the 8 limitations on this vintage are marked high severity by the pipeline itself.
- 3 of the 35 states carry no composite. Coverage falls below the declared threshold, the cause is published per state, and the value is left absent rather than imputed.
- 1 of the 5 declared weights falls outside its own confidence interval, estimated over 32 complete cases with 4000 resamples. The page reports the realised influence, not the intention.
- The release blocks itself. This page is regenerated from the index and the build fails when a figure here stops matching the asset it came from, so a stale number cannot reach a reader by being forgotten.
Assets: data/qesis_v8.json, concordance errata and uncertainty ledger, vintage v9.1 (2026-09-17). Provenance chain and the full validation record travel with this page as machine readable metadata.
How a number gets from a public registry onto this page
Four runtimes, one gate, and a cleared vintage that this page carries rather than requests. Intake is deliberately permissive and emission is deliberately strict, so everything ambiguous is settled in one place instead of in every place. The three feeds that stop at the gate in the figure are not decoration: they are the 3 states whose coverage fell below the declared threshold, withheld and published as a gap rather than filled in.
- continuous intakeSCOUT
Reads the declared source registry for every axis and records the vintage, the licence and the coverage at the point of collection, so a figure can always be traced back to the release it came from.
- the compliance gateSENTINEL
One gate, three verdicts, and nothing reaches the index or this page without passing it. Coverage below the declared threshold is withheld and published as a gap rather than filled in, which is why 3 of the 35 states carry no composite and each says why.
- equifinality engineANALYST
Reduces the truth table of the six measured conditions to the smallest set of combinations that are each sufficient on their own, publishes the combinations it cannot rely on beside the ones it can, and states what would falsify the result.
- the island builderARCHITECT
Writes the cleared vintage into a hash chained record this page carries with it, so a reader fetches nothing at run time and a figure that stops agreeing with its source fails the build instead of reaching a screen.
Mandates summarised from sovereign-infra/agents, which remain authoritative. The registry is closed at six runtimes; the two not drawn are HERALD, which publishes only from cleared material, and COUNSEL, which owns money and law. Counts derived at build time from data/qesis_v8.json.
The constraint, and why it is a reading rather than a variable
Algorithmic independence, hyperscale efficiency and ecological sustainability do not fit in one state at once. That is the claim the work is built on, and it is carried here as an interpretation of the measurements, never as a term inside the model. The reason is visible in the figure: one measured condition is claimed by two corners at the same time, and two conditions belong to no corner at all. A constraint whose corners cannot be told apart by the data cannot also be a variable in the model that data feeds.
Corner assignment read from the served statement that governs how this claim may be used. No result on this page is described as a corner of the constraint, because the pipeline forbids it.
What the measurements do price is the cost of the trade. Across 32 states the six axes are weakly entangled at 0.124. Restricted to the states that import their energy, their minerals and their hyperscalers, entanglement rises to 0.176 over 26 states. The constraint is cheap where the substrate is domestic and expensive where it is imported, and the gap between those two numbers is the whole of it.
That is also the only form of the constraint the measurements support you handling directly. There is no calibrated function here that makes a corner collapse when you raise the other two, and staging one would be inventing a law this model does not carry. What was measured is entanglement: pick an axis and see what is observed moving with it, across all 32 states and again inside the import core.
- CSE+0.070+0.089
- REE-0.259-0.160
- FPE-0.211-0.490
- ODI-0.179-0.509
- ESE+0.360+0.576
- WSE+0.070+0.089
- REE+0.114+0.138
- FPE-0.095-0.105
- ODI-0.072-0.148
- ESE+0.137+0.099
- WSE-0.259-0.160
- CSE+0.114+0.138
- FPE+0.197+0.295
- ODI+0.245+0.387
- ESE-0.281-0.304
- WSE-0.211-0.490
- CSE-0.095-0.105
- REE+0.197+0.295
- ODI+0.783+0.755
- ESE-0.385-0.445
- WSE-0.179-0.509
- CSE-0.072-0.148
- REE+0.245+0.387
- FPE+0.783+0.755
- ESE-0.423-0.485
- WSE+0.360+0.576
- CSE+0.137+0.099
- REE-0.281-0.304
- FPE-0.385-0.445
- ODI-0.423-0.485
Correlation stated as correlation, not as cause. Bars run from the centre, right for positive and left for negative, on the same scale in both columns. Asset: data/qesis_v8.json, coupling block, vintage v9.1 (2026-09-17).
Ten routes to the same failure, not one
There is no single road to high substrate vulnerability. The analysis returns 10 combinations of conditions that are each sufficient on their own. 6 of them clear the working conventions and are relied on; 3 sit below the specificity convention and are shown without being relied on; 1 fall far enough that they are published for completeness and carry no weight. Nothing is dropped, because a route removed from a public surface is a route nobody can check.
| Route | Water stress | Cable exposure | Minerals | Operator | Sovereignty | Cyber gap | States | Verdict |
|---|---|---|---|---|---|---|---|---|
| Route 1 | 1 | Not relied on | ||||||
Present in this route: Submarine cable exposure, Critical minerals dependence, Hyperscaler concentration, Cybersecurity commitment gap. Absent in this route: Water stress, Effective sovereignty not held. States on this route: 1. Agreement 0.8841. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.7455. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1043. Verdict. This route sits below the PRI working convention of 0.75. It is published rather than dropped, and no claim rests on it alone. | ||||||||
| Route 2 | 1 | Not relied on | ||||||
Present in this route: Water stress, Critical minerals dependence, Effective sovereignty not held. Absent in this route: Submarine cable exposure, Hyperscaler concentration, Cybersecurity commitment gap. States on this route: 1. Agreement 0.8613. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.6727. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1143. Verdict. This route sits below the PRI working convention of 0.75. It is published rather than dropped, and no claim rests on it alone. | ||||||||
| Route 3 | 3 | Relied on | ||||||
Present in this route: Water stress, Critical minerals dependence, Effective sovereignty not held, Cybersecurity commitment gap. Absent in this route: Submarine cable exposure, Hyperscaler concentration. States on this route: 3. Agreement 0.9347. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.8499. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1586. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
| Route 4 | 2 | Relied on | ||||||
Present in this route: Water stress, Critical minerals dependence, Hyperscaler concentration, Effective sovereignty not held, Cybersecurity commitment gap. Absent in this route: Submarine cable exposure. States on this route: 2. Agreement 0.9535. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.9187. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1157. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
| Route 5 | 1 | Relied on | ||||||
Present in this route: Water stress, Submarine cable exposure. Absent in this route: Critical minerals dependence, Hyperscaler concentration, Effective sovereignty not held, Cybersecurity commitment gap. States on this route: 1. Agreement 0.9322. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.8534. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1090. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
| Route 6 | 1 | Set aside | ||||||
Present in this route: Water stress, Submarine cable exposure, Cybersecurity commitment gap. Absent in this route: Critical minerals dependence, Hyperscaler concentration, Effective sovereignty not held. States on this route: 1. Agreement 0.8273. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.4910. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1016. Verdict. This route falls below PRI 0.50, which means it is close to equally consistent with the opposite outcome. It is published for completeness and carries no weight. | ||||||||
| Route 7 | 1 | Relied on | ||||||
Present in this route: Water stress, Submarine cable exposure, Critical minerals dependence. Absent in this route: Hyperscaler concentration, Effective sovereignty not held, Cybersecurity commitment gap. States on this route: 1. Agreement 0.9179. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.8334. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1477. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
| Route 8 | 2 | Not relied on | ||||||
Present in this route: Water stress, Submarine cable exposure, Critical minerals dependence, Effective sovereignty not held. Absent in this route: Hyperscaler concentration, Cybersecurity commitment gap. States on this route: 2. Agreement 0.8426. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.6488. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1218. Verdict. This route sits below the PRI working convention of 0.75. It is published rather than dropped, and no claim rests on it alone. | ||||||||
| Route 9 | 1 | Relied on | ||||||
Present in this route: Water stress, Submarine cable exposure, Critical minerals dependence, Hyperscaler concentration, Cybersecurity commitment gap. Absent in this route: Effective sovereignty not held. States on this route: 1. Agreement 0.9568. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.9179. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1193. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
| Route 10 | 2 | Relied on | ||||||
Present in this route: Water stress, Submarine cable exposure, Critical minerals dependence, Hyperscaler concentration, Effective sovereignty not held. Absent in this route: Cybersecurity commitment gap. States on this route: 2. Agreement 0.9602. Among the states that combine these conditions in this way, that is the share which also carry high substrate vulnerability, on a zero to one scale. Specificity 0.9371. How much of that agreement is specific to high vulnerability rather than equally true of its opposite. The working convention is 0.75, and 0.50 is the point below which a route is set aside. Share of the outcome this route accounts for: 0.1232. Verdict. This route clears the working conventions the analysis publishes. | ||||||||
What each column measures
- Water stress. axis WSE, Water Stress Exposure. Aqueduct 4.0 baseline and SSP3.
- Submarine cable exposure. axis CSE, Cable Stress Exposure. EMODnet density and the cable registry.
- Critical minerals dependence. axis REE, Rare Earth Element Stress. USGS and CRMA.
- Hyperscaler concentration. axis ODI, Operator Dependency Index. Concentration over facility counts.
- Effective sovereignty not held. derived. Effective Sovereignty Extended is a jurisdictional and key sovereignty measure over cloud infrastructure: the mean of key sovereignty, supply chain integrity, resource sovereignty, restoration sovereignty and the jurisdictional contract average, on a zero to one scale, inverted here so the condition reads as the share a state does not hold. Five of the 35 states are collected from the pilot audit and thirty are inferred from regulatory framework assessment, and any claim resting on this condition carries that split. Codebook.
- Cybersecurity commitment gap. derived. The ITU Global Cybersecurity Index, 2024 fifth edition, scores state level legal, technical and organisational cybersecurity commitments out of 100, and the condition is the remaining gap. Coverage 33 of the 35 states. Codebook.
Across all 10 routes together, agreement 0.9048 and coverage 0.5807 over n=32. Coverage below one is deliberate: the solution explains the cases it covers and declines the rest. Asset: data/qesis_v8.json, solution block, vintage v9.1 (2026-09-17).
What is not claimed. No condition is publishable as necessary at this vintage. None reaches the conjunctive bar the pipeline declares, and one that would have passed a weaker single test is refused by it. The full verdict travels with this page as metadata.
The cable network tolerates random loss and fails under an aimed one
An independent rebuild of the topology reproduces the published articulation set exactly, 88 of 88, which is what makes the rest of this block citable. Over 912 cities and 343 cables, removing chokepoints at random costs little. Removing them in order of how much traffic they carry is a different curve entirely, and the first twelve removals are cheap. The thirteenth is not.
Those two thresholds are different quantities and are published side by side so neither is quoted as the other. The single step severance is the largest one step loss and it arrives at removal 13. Half collapse arrives later, at removal 19, when the connected core first falls below half its baseline. The node at the tip in this release is Porthcurno, United Kingdom, and which city sits there is a property of the release rather than of the finding.
The removal order, as the pipeline computed it
- 1Mazara del Vallo, Italy7 cut off
- 2Fortaleza, Brazil1 cut off
- 3Penmarch, France1 cut off
- 4Ajigaura, Japan1 cut off
- 5Sesimbra, Portugal1 cut off
- 6Norden, Germany1 cut off
- 7Bude, United Kingdom2 cut off
- 8Makaha, Hawaii, United States13 cut off
- 9Melkbosstrand, South Africa1 cut off
- 10Shima, Japan1 cut off
- 11Mtunzini, South Africa1 cut off
- 12Marseille, France1 cut off
- 13Porthcurno, United Kingdom278 cut off
Assets: data/qesis_percolation.json for every published figure, data/axes/cse_percolation.json for the per step curve. The build asserts the two agree at the two published points before this page is written.
What would refute all of this
The fragmentation finding fails if an independent topology of comparable coverage produces an aimed removal curve whose largest single step loss falls inside the range of its own random curve, or if the 88 articulation cities fail to reproduce from that topology. A different city at the tip does not refute it. The equifinality result fails if the solution does not survive the declared anchor sensitivity, which is why both anchor regimes are published and their disagreement is carried as data rather than settled by preference.
Standing order: every headline finding states what would falsify it. A finding published without one is not published here.
Point your own systems at it, rather than at this page
A report nobody can act on is the failure mode this instrument was built to avoid, so the surface is not the product. The same values this page renders are served as tools an analyst or an agent calls directly: qesis_get_coupling, qesis_get_pathways, qesis_get_country, qesis_rank_countries and qesis_get_integrity, the last of which answers which generation you are reading and whether its chain verifies. Every figure above is derived at build time from the assets named in the source lines, so this page cannot say something those tools do not.
The deployment plane serves what was promoted and the local plane reads the working tree, and a reader is always told which one they hold. The full country dashboard carries the per state profiles and the index summary carries the ranking and the published gaps.
Nothing you do here is logged, no cookie is set and there is no backend behind this page. The validation record travels with it as metadata rather than as a request to a server.