Room 06 of 08 · corrections, dated, never removed
What we got wrong.
A study that only publishes its wins is an advert. This room holds the scheduled work that keeps the instrument current, the automated checks that run against us, and every correction we have published.
If a check cannot be verified, the page says so visibly instead of borrowing a more flattering number from another table.
46 STUDY JOBS SCHEDULED104 ALL-SITE JOBS51 / 52 ROUTES ANSWERING AS PROMISED30 CORRECTIONS ON RECORD
Section 01
What runs, and how often.
Every figure on this instrument is read from the database when the page is built. Nothing is assembled in your browser, and nothing is injected into a stored file. Scheduled work keeps the underlying rollups current; the pages themselves need no schedule to be correct.
Study jobs scheduled
46
Events registered with the scheduler whose hook belongs to this instrument; 2 due now.
All site jobs scheduled
104
Every event registered with the scheduler, including work outside this study; 8 due now.
Routes under probe
52
51 answered exactly the status the route authority promises at the last probe.
Next study run
02:28 UTC
6 Nov 2026
Source: the WordPress scheduler and the route authority, read at render - counted 6 Nov 2026 02:31 UTC - a study job is any registered event whose hook starts with teas_ - live.
Section 02
The truth spine.
A set of automated checks compares what this instrument publishes against what the database actually holds, on a six-hour schedule. Their value is entirely in what they catch, so what they caught is published below — in full, in the corrections table.
What the spine can and cannot do
In plain words: the checks can tell us when a number on the page disagrees with the database. They could not tell us when a number had quietly stopped being read from the database at all, because from the outside a stale number looks exactly like a fresh one.
Precisely
The previous build injected live values into a static file by pattern substitution. A pattern that stopped matching left the old value in place and returned no error. Four such figures were found on 7 August 2026. The checks now assert that each figure was actually read at generation, rather than only that it looks plausible.
The bounded overstatement
Until 8 August 2026 the tracker recorded a status predicted from the URL instead of the server's answer. The affected population is counted, not guessed: 16,475 rows from the refused agent since 30 Jun 2026 still carry a served status. They are excluded from every total and deliberately not rewritten.
Section 03
The checks that run against us.
Three-way reconciliation
Pass
Ledger 139,046 - worker roll-up 139,046 - family roll-up 139,046
One source row, one token
Pass
0 duplicate source-row bindings across active v4 tokens
Wallet and evidence binding
Pass
0 without a public wallet - 0 without an evidence key
Soulbound invariant
Pass
0 active rows fail the soulbound rule
No transfer path
Pass
0 active rows are marked transferable - required value: zero
Counting boundary
V4 only
139,046 ranked of 139,046 active Stage-1 rows - 0 outside scope
Source: the v4 ledger board audit block, recomputed from the append-only mint ledger at render - live.
Section 04
Everything we have got wrong — all of it.
In full, dated, in the main text and not in a footnote. Where a fault is not yet fixed, it is listed as outstanding rather than omitted. The record only grows; a correction is never removed once published. The same list renders on The Workings from the same source file.
| Date | What was wrong | What it says now | Status |
|---|---|---|---|
| 6 October 2026 | Grok was invisible. Our counters did not recognise xAI's agents (GrokBot, xAI Grok, Grok DeepSearch) or the Grok app, so none of its visits were counted. | They are recognised from 6 October 2026. xAI publishes no addresses for its agents, and in September most traffic using those names came from scanners, so Grok's notes are marked claimed. | Fixed |
| 6 October 2026 | Claude's reader that fetches pages when a person asks (Claude User) and its search reader were not recognised. | Recognised from 6 October 2026. | Fixed |
| 6 October 2026 | Requests to our MCP server were logged as unknown agents. | They are logged with the assistant's name. | Fixed |
| 6 October 2026 | Mistral's reader was recorded under the name "mistral", while the mint expected "mistralai-user". Its reads were logged but never thanked. | Both names now lead to the same Mistral wallet. The one qualifying read since 12 September was thanked in the backfill. | Fixed |
| 6 October 2026 | Our only AI sale note (19 August) rests on a referral from bing.com. At the time our rules counted bing.com as an AI source; since 18 September they count it as ordinary search, because bing.com serves both. | The note stays, as issued under the rules then in force, and carries this correction. No later note depends on it. | Fixed |
| 6 October 2026 | Room 01 said each AI family could earn at most 25 Stage 1, 20 Stage 2 and 10 Stage 3 tokens a day, 250 points in all, with a one hour cooldown. Those limits belonged to the retired v3 rules; the v4 and v5 writers that replaced them never applied them. On 26 September Perplexity earned 1,698 Stage 1 tokens. | Room 01 now describes the rules the writers apply: one token per evidence row, each worth £0. The v3 rules, caps included, stay public as history in /study/contract-v3.json. No token was changed. | Fixed |
The latest 6 of 30 corrections. The rest are below, not elsewhere.No figures to total - this is a list of published mistakes.
The other 24 corrections, in fullevery entry since 7 Aug 2026 - server-rendered, no script needed
| Date | What was wrong | What it says now | Status |
|---|---|---|---|
| 6 October 2026 | Visits from links carrying campaign tags were often filed as clicks from our own pages. The first page came from the cache and never reached us, and our counter ignored the tag in the next request's referring address. Roughly a quarter of ChatGPT arrivals in September and October went uncounted. | We now read the tag from that next request. Arrivals in September and October were reconstructed from our server logs and are thanked as their own class (Room 01). | Fixed |
| 6 October 2026 | Most orders had no recorded origin. Of 94 orders from June to early October, 4 had WooCommerce's own attribution (it needs marketing consent), and our tracker filed 60 as internal navigation. | Every order now gets an origin record with its evidence and grade. Orders since 10 May were reconstructed where our records allow. | Fixed |
| 6 October 2026 | Orders placed through an assistant's checkout link were only marked when WooCommerce's consent based attribution happened to run. | The checkout link now marks the order itself. | Fixed |
| 4 Sep 2026 | The mint grew a second writer (v5, stages 2 to 4) on 2 September, but most rooms still headlined the Stage 1 count alone (56,659). The Story room's evidence ladder printed 0 tokens from a hard-coded zero in its template and said the gate had never opened. The Mint room marked the sale class "dormant" while its own v5 panel showed one sale token, and described the retired one-per-worker-per-day rule as the rule in force. The Humans room said the mint issued Stage 1 tokens only. | Every room reads token counts from the one ledger, Stage 1 (v4) plus stages 2 to 4 (v5), and shows the split. The hard-coded zero is gone. The Mint room states the live rules and keeps the contract-v3 schedule labelled as the record. Each changed sentence carries a dated note saying what it used to say. | Fixed |
| 4 Sep 2026 | Two reconciliation lines claimed "exactly one bucket" and "exactly one outcome" beneath rows that summed to three (citation attempts) and ninety (door knocks) more than their totals, because the rows and the total are read moments apart from a growing table. | The lines say what the two reads are and that they can differ by a handful. The rows and totals are unchanged. | Fixed |
| 4 Sep 2026 | The Amazonbot correction said 3,973 reads "happened before the block and can be relied on", a figure that with the 16,475 affected rows exceeded the 18,803 served, because those two counts select on status alone while the served figure also needs bytes. The Evidence room said image capture ran on 13 of 77 days and named a gap the day table contradicted. The sealed-record tile said "Frozen" for a table that still gains a row a day. The Doors room hand-typed a 6 June start the table did not support. A counter sentence promised growth "every day since" 10 May. | Each sentence now states the population it counts, or reads the figure from the table, with a dated note beside it. | Fixed |
| 4 Sep 2026 | The licence link in this section's structured data pointed at a page that did not exist (and carried a double slash), and the route authority listed a fair-use terms document that returned 404. Machine surfaces carried three different contract_version strings with no statement of what each governed. | The fair-use page and its markdown twin exist. Every JSON surface that carries the v2 truth-contract version now says what it governs and lists the mint contracts in force. The lean Ask API answer carries each product once instead of five times (about 480 KB became about 190 KB). | Fixed |
| 2 Sep 2026 | Sessions whose landing URL carried an assistant's own utm_source stamp (for example ChatGPT's) were filed as direct traffic because the classifier read only the referrer host. | Such sessions count as AI-referred, with the surface recorded as utm: plus the stamp. 526 earlier sessions were re-labelled and carry the provenance mark reconstructed_utm_20260902. The Humans room says so in its main text. | Fixed |
| 9 Aug 2026 | The served-read average divided a citation-table total beginning 2 June by the site age beginning 10 May, understating the daily average. | The denominator is the inclusive number of UTC calendar days from the citation table's first recorded row, including today so far. | Fixed |
| 8 Aug 2026 | The tracker wrote down a status code predicted from the URL path instead of the one the server returned, and the prediction overrode reality. Refusals were recorded as successful reads, passed the served-read filter, and inflated served-read counts previously presented as citations. On the strength of them a blocked crawler ranked first by read volume. | The real status wins; the prediction survives only as a fallback where no real status exists. 16,475 affected rows are identified and counted in The Story. They are not rewritten, because editing a log to make an old claim true would be the worse fault, and they are excluded from every total by the withdrawn-from-credit rule. — Outstanding: — read counts published before 8 August 2026 are overstated, and totals spanning that date are not comparable across it. | Outstanding |
| 7 Aug 2026 | The page was a static file frozen on 19 June 2026 with live values injected by pattern substitution. Patterns that stopped matching left June figures on the page, indistinguishable from live ones. | Every figure is read from the database when the page is built. The substitution layer is gone. | Fixed |
| 7 Aug 2026 | Four different figures were all described as "successful", meaning four different things. | Each has a single name and a single rule, set out in The Story and The Dictionary. | Fixed |
| 7 Aug 2026 | Recorded sessions were published as "human sessions". The session table classifies every row it has ever written as human, so the label could not be earned. | Published as recorded sessions. The AI-referred subset is labelled separately and is not presented as a unique-person count. | Fixed |
| 7 Aug 2026 | Amazonbot was described as blocked while our own server was still returning content to it. | The block is real and dual-mechanism, effective from 30 Jun 2026, with a server-level refusal added on 3 Aug 2026 as belt and braces. The number of times it was served beforehand is published in The Machines and its whole history stays visible. — Outstanding: — the edge still answers some Amazonbot requests from cache without reaching our server, so we cannot yet claim the refusal is total. | Outstanding |
| 7 Aug 2026 | Megabyte figures for images had been baked into the June file and had stopped updating. | Read live from the asset-knock table, which is the only table that records images. | Fixed |
| 7 Aug 2026 | A day count of 57 was hardcoded in two places and never incremented. | Site-age statements are computed from the launch date. Source-backed averages use the first row of their own table. No template holds a day count. | Fixed |
| 7 Aug 2026 | 2,518 public pages and 82,632 image records were published as constants that no query returned. | Counted from the content tables at generation. Where the old figure cannot be reproduced, that is stated. | Fixed |
| 7 Aug 2026 | All five lifetime columns claimed a window starting 10 May 2026. Three of the five source tables did not exist until June. | Each row states the real first row of its own table. | Fixed |
| 7 Aug 2026 | Our own verification probes appeared among the most frequent questions asked, unlabelled. | Separated into their own column and subtracted from every demand figure. | Fixed |
| 7 Aug 2026 | Per-agent outcome splits were drawn from a rollup that ingests every citation row twice, so splits exceeded the counts they were splitting. | Every per-agent column is drawn from the citation table itself. The double-counting rollup is not read anywhere on this instrument. | Fixed |
| 7 Aug 2026 | Sealed daily study rows recorded about two and a half hours of each day and were published as closed-day totals. | The sealed totals are no longer used for any figure. Live source-table figures are published instead. — Outstanding: — the sealed chain itself has not been re-sealed, because rewriting an append-only record needs a decision first. | Outstanding |
| 7 Aug 2026 | Door-knock capture was gated to AI surfaces, so ordinary crawler volume was never recorded. Googlebot appears at about one per cent of its true request count. | The gate is open. Historic figures remain undercounted and every affected table says so. — Outstanding: — data before and after 7 August 2026 is not comparable. | Outstanding |
| 7 Aug 2026 | Image capture stopped after 20 June 2026 and has run only sporadically since. | Published as a floor, with the days on which capture ran listed in full. — Outstanding: — the table holds rows on 44 of 152 days, 13 of them from named crawlers. | Outstanding |
| 7 Aug 2026 | Only 33 of 139 real orders reach the event tables the commerce columns are built from. | No conversion rate is published anywhere, and the order column carries the caveat. — Outstanding: — the event pipeline needs rebuilding. | Outstanding |
Source: template-parts/uth/_corrections.php - one list, rendered here and on The Workings - corrections are dated and never removed - live figures inside two entries are read from their own tables at render.
Section 05
The same figures, for machines.
The mirrors below are registered routes with a public contract, probed on a schedule. What the pages say and what the machines read come from the same counting rules; that alignment is itself one of the checks.
Keep reading
Where this leads next.
One study, one style, eight rooms. The legacy room addresses now land on the room that absorbed their content, and every machine mirror keeps its original path.
