How this was made. The version number counts drafts of the text. It does not measure the inquiry behind it, which has run over days and across several AI systems, with argument between those systems and within them, directed, refused and repeatedly redirected by the author. The source material was AI-generated, and then adversarially and iteratively refined across a range of tools — systems built by different companies in different jurisdictions, set against each other and against the author. No one of them produced this text, and no one of them reviewed it alone. The plurality is deliberate rather than incidental. A single model carries a single set of priors about which sources are authoritative, and this series argues that an evidence base narrowed in exactly that way is how a contested question comes to look settled. Using one model to investigate that claim would have been the claim refuting itself. To name a single model on it would credit that model with work that was neither its own nor done in a single pass. The plurality was also necessary, and the record should say why. In drafting, the assisting model repeatedly led with United States institutional sources — a national laboratory, an industry association, a market study nineteen years old — and presented conclusions drawn from them as the state of knowledge. On one occasion European measured data contradicting those conclusions was present in the same research return and was placed below them. Framings were proposed that would have argued against this series’ own position using that evidence base, and offered as rigour. Each was refused by the author and the material rebuilt. That is the mechanism these documents describe, occurring in their own making, and it is recorded because a series arguing that evidence bases narrow without anyone deciding to narrow them cannot credibly claim its own production was exempt. The framing, the corrections and the judgements are the author’s, and so are the errors. How this site is written sets out what is declared on every piece, who checks it, and where the per-piece record lives.
What this publication does not claim, and what is outstanding against it in the register.
Asserted, not yet measured: this publication.
A question this rests on is parked: What secession premium will members bear?
We publish no pricing and no affordability claim, and we do not assert that members value the secession guarantee.
What this publication rests on, and how solid each part of it is. What nobody has measured yet is an essay, and this page describes it as one.
This publication cites outside sources, and they are listed below — each with what it supports, and with what it does not support. That second column is the one that matters: the common failure is not a fabricated source, it is a real source stretched past its finding.
report
Supports. Cited as describing “the estate as it was built between roughly 2010 and 2024” via power usage effectiveness (PUE) — used to characterise the limits of existing data-centre measurement.
Does not support. “PUE… says nothing about whether the work being done is useful, nothing about the cost of the software arrangement around it, and nothing about the governance properties this series is concerned with”; figures are operator-reported and unaudited, “one scheme discarding a portion of its returns as malformed and the most recent report flagging two of its own indicators as misaligned with reality.”
⚠️ Retrieval. Operator-reported and unaudited per the document’s own characterisation; figures deliberately not reproduced.
Supports. Cited as modelling “the American estate over the same era” — used with the EU report to characterise the limits of existing measurement, and later as a figure conflicting with a grid-water number used elsewhere in this series.
Does not support. The American figures “are simulated — their authors state that the simulations ‘assume systems are commissioned and operate as designed. This is rarely the case.’”
⚠️ Retrieval. Simulated, not measured, data by the report’s own admission; figures deliberately not reproduced.
report
Supports. The origin of the claim that most servers sit in small rooms.
Does not support. The study predates cloud computing, so the claim cannot be read in the present tense.
⚠️ Retrieval. A pre-cloud market study, and the estate it describes no longer exists in that shape.
paper
Supports. Named as the origin of “a grid-water figure” used elsewhere in this series.
Does not support. The figure travelled through two intermediaries from a working paper, and the American laboratory’s own figure for the same quantity is materially different. It is not reliable enough to carry an argument.
⚠️ Retrieval. Reached only via two intermediaries, and it conflicts with LBNL’s figure for the same quantity.
paper
Supports. The source of a manufacturing-carbon share, and that share belongs to consumer handsets.
Does not support. The same authors give a substantially lower range for servers, so the handset figure cannot be applied to server hardware.
⚠️ Retrieval. The handset and server figures are easily conflated, and the difference between them is material.
paper
Supports. Cited: “built a cluster from second-hand handsets and found it better on carbon per unit of work than a cloud instance over three years” — thin evidence that smaller, cheaper, abundant hardware could invert current computing economics.
Does not support. “It is a single comparison on three workloads, and the same literature finds that permanently powered devices behave differently from the phones it used — so it establishes that the question is open, and not that the answer is known.”
Supports. Cited again: “no measured reduction in failure rates” — a caution about approval as a mechanism, supporting the claim that governance cost may be the binding constraint and is unmeasured.
Does not support. Not a measurement of AI-agent governance cost specifically; the piece states plainly “nobody has published a figure” for identity infrastructure, policy enforcement, audit storage, or human approval time — this is the nearest available evidence, not that measurement itself.
⚠️ Retrieval. Same source referenced in “How much it may do unsupervised”, where it is disclosed as vendor-published, self-reported, and cross-domain (software delivery, not AI agents).
framework
Supports. Cited: “records that onshore hosting with a foreign-headquartered provider still leaves configuration and hypervisor access abroad” — one of three observations bearing on whether sovereign-arrangement properties can be assembled on infrastructure an institution does not govern.
Does not support. Presented as one of “three observations, not a proof” — “consistent with the current path being unable to deliver the arrangement, and… also consistent with it delivering a weaker version that many institutions would accept.”
evidence
Supports. Quoted again: “Non, je ne peux pas le garantir” (adding it had never happened) — one of three observations in the fourth fork about whether sovereign arrangements are achievable at all on vendor infrastructure.
Does not support. As in “Who actually holds the controls”: establishes structural legal exposure, not that any transfer has actually occurred.
⚠️ Retrieval. This document refers to “a vendor’s own legal officer” and does not name him. The name is supplied here because a sources page exists to identify what a document leaned on, and the same testimony is cited by name in “Who actually holds the controls”. The testimony is a matter of public record.
other
Supports. Cited: “runs its control plane locally and requires the vendor’s approval, a current support agreement and an annual renewal for the institution to be permitted to run it at all” — first of three observations suggesting current infrastructure may be structurally unable to deliver full institutional authority.
Does not support. One of “three observations, not a proof”; not claimed to generalise to all vendor sovereign-cloud offerings.
⚠️ Retrieval. The document does not name the vendor or the product, and this entry does not supply them. The description matches material quoted by name in “Who actually holds the controls”, but that is an inference from similarity rather than a citation, and it is recorded as one.
Foundational documents. These are positions this project has taken, not findings.
| Record | What it is | Status |
|---|---|---|
CON-13 |
Claims carry evidence | draft v0.1 |
🔴 This publication is tagged requires-evidence in the corpus and references none. The tag is the project’s own record that this document is the kind that needs measurement behind it. That measurement is not yet recorded here, and this page says so rather than leaving the tag to be found by someone reading the frontmatter.
None. This publication references no evidence record. That is the correct description of what it is rather than a gap: it is an essay, reasoning from the foundational documents above rather than reporting a measurement. Where it states a number, that number is marked in the text as what it is.
| Record | What it is | Status |
|---|---|---|
CON-10 |
Bounded reversibility, honestly stated | draft v0.1 |
FIG-31 |
Four forks, and the observable that tells you which one you are on | drawn for this publication · figures/FIG-31.svg |
Open questions this publication is known to be exposed to. They are recorded rather than resolved, and the set card counts them.
OQ-02Generated from the corpus, not written by hand: this page cannot claim a source the corpus does not hold, and it changes when the records do.
Alongside: the publication · questions and answers