The claims behind this work, the source each was checked against, and what the check found, including the ones we got wrong first and the ones we still cannot settle. The correction mechanism applied to our own file.
Sociable Systems · Working method
The Verification Ledger
The claims behind our South Africa data centre work, the source each was checked against, and what the check found, including the claims we got wrong first and the ones we still cannot settle.
Most commentary on this buildout is assembled from search snippets and second-hand summaries, and a meaningful share of it does not survive contact with the primary record. This page is the correction mechanism applied to our own file. It is updated as items close, and corrections are logged rather than silently folded in.
4verified against primary sources
2corrected after checking
3still open on our own thesis
8named-firm claims held unpublished
Verified
Checked directly against the named source: the document itself, not a summary of it.
01Verifiedchecked against Daily Maverick, 24 Aug 2026 · 2026-08-28
The Daily Maverick article resolves, and the “confirmed after approval” wording appears in it verbatim, in the authors’ own account of asking about water consumption.
One summary in circulation cited a URL that does not resolve. The article is real; the citation was not. Load-bearing for the public-register essay and the Tuesday episode.
02Verifiedchecked against tribunal record / Daily Maverick · 2026-08-28
Cape Town’s Municipal Planning Tribunal approved the King Air Industria land-use application on 14 July 2026, one member dissenting for lack of information.
The exact vote split and the appointed town planner remain unconfirmed (item 07). The fact of the approval, the date, and the dissent are settled.
03Verifiedchecked against publicly hosted application documents · 2026-08-28
A s24G rectification for unlawfully commenced generator-capacity infrastructure at a data centre exists and is publicly documented (Erf 358, 401 Old Pretoria Road, Midrand, plus a further expansion).
Two conflicting reference numbers circulate for what may be two different processes; until the register resolves which attaches, we cite that the rectification happened, not its number.
04Verifiedchecked against the submission itself · 2026-08-28
The SAHRC submission is dated 24 August 2026, with Housing Assembly and Foxglove among the signatories, and makes four demands as characterised.
The four demands read as a scope of work: a participatory national inquiry, an independent cost/benefit assessment, a pause, and an independent monitoring panel.
Corrected
Claims we circulated internally, then found to be wrong when checked. Logged here so the correction is part of the record.
05Correctedchecked against EirGrid / commission decisions · 2026-08-28
Ireland’s connection moratorium ran from 2021 and reopened in December 2025 on new terms: own generation, with 80% of new domestic renewables within six years. The Dublin region remains constrained to at least 2028.
The drafts had the timing and the reopening terms wrong. Stated accurately, the comparator is stronger, not weaker: a jurisdiction can pause, set conditions, and remain a global hub.
06Correctedchecked against NSW planning framework · 2026-08-28
New South Wales has no bespoke data-centre SIA rule. It pulls large data centres into the state-significant-development track on capital investment value, and that track carries social impact assessment as standard.
The drafts asserted a dedicated rule that does not exist. The actual mechanism is more interesting: scale alone triggers the social assessment, with no special category required.
Open, including on our own thesis
These are marked as inferences wherever they appear in our published text. They are listed here precisely because two of them sit under arguments we are making.
07Opento be checked against Cape Town Municipal Planning By-law, direct read
Whether the Site Development Plan stage carries public-participation standing.
The highest-value open item in the file. It is the load-bearing claim under the deferral argument, it is currently an inference rather than a finding, and the published arc marks it as an inference in body text rather than asserting it. Until the by-law is read directly, nothing we publish treats it as settled.
08Opento be checked against MPT record
The exact tribunal vote split (reported as 4–1) and the town planner’s name on the Equinix application.
Not used until confirmed, in any register.
09Opento be checked against GDARD register
Which of two conflicting reference numbers attaches to the Midrand rectification.
Draft 2 does not go out until this resolves. The fact of the rectification is verified; its file number is not.
Held: named-firm claims, unpublished
Multiple research passes produced firm and practitioner names attached to specific projects. Some are real. Some are single-source and may be invented. Until each survives a register check, it stays here and out of everything public, because the outreach that follows this work goes to exactly these people, and a name published wrongly converts a warm introduction into a defensive one.
10Held
A named environmental assessment practitioner for the King Air Industria application, asserted by three of seven research passes.
Three passes may share one bad source. Register check pending.
11Held
A firm’s authorship of a well-known Cape Town bar process, plus unverified “confidential data centre projects” line items.
The first half is likely true and checkable; the second requires a CV we have not seen.
12–14Held
Three further firm–project attachments, each from a single research pass.
Single-source. One is specific enough to be checkable from the tribunal record alone.
15Held
A firm’s published 2026 argument that South Africa’s framework does not fit hyperscale.
If true, this is the closest ideological neighbour to our position and a competitor for the same ground. Verification has priority.
16Held
Conflicting attributions of a firm’s practitioner of record.
Both may be true on different files. Unresolved, so neither is named.
17Held
A scatter of firm names with no project attached.
List-padding. Ignored until a named project attaches.
How this works
Every claim is checked against the document it came from, not a summary or a search snippet.
Corrections are logged, not folded in silently. The two comparator errors above were ours, and they are the reason this page exists.
Anything still open is marked as an inference in the text that uses it, or the text waits.
Named firms and practitioners stay out of public text until a register check closes the question.
The ledger updates as items close. New claims enter at the bottom and work up.
The standing rule
Nothing is published with an unverified named firm in it. A wrong claim costs the writer a correction; a wrong name costs the person named. The second is not ours to spend.
