A repository of consumer maintenance, repair and troubleshooting procedures, where the data is the product and the web page is a view of it.
What this is
Each procedure is stored as a structured record: steps, hazards, tools, terminology, sources and specifications, each with its own verification state. The page you read is rendered from that record, and so is the printable version, the structured data and the feed. Building it once as data is what makes it possible to check the whole corpus at once.
What is different about it
Two things, and neither of them is volume.
The first is controlled language. Procedures are written to ASD-STE100 Simplified Technical English, the standard used for aircraft maintenance documentation. One term per item, held throughout. Short imperative sentences. Conditions before actions. It is designed to be unambiguous to a reader who is tired, stressed, or reading English as a second language.
The second is traceable specification. Every number is traced to a manufacturer document, a published standard or a government safety authority, and the source is named with the date it was read. Where a number cannot be traced, the page says so in the place the number would have been, rather than printing a plausible guess.
Who writes it
An editorial system does, and it says so. There is no author persona here, and none has been invented — no fictional expert, no stock photograph, no claimed credential. Trust is meant to rest on the published method and the public correction record, both of which you can check, rather than on a biography, which you cannot.
The method is described in full on the standards page. That page is the honest answer to "why should I believe this".
When it is wrong
It will sometimes be wrong. The response to that is built in rather than bolted on: a public corrections log that is never edited, a version number on every page, verification before any correction is applied, and immediate removal of a page whose specification is confirmed wrong. A page that is down costs nothing.
If you find an error, comment on the page or write to the corrections address on the contact page. A correction confirmed by a first-party source is applied and logged.
How it runs
Scheduled routines find topics, research them against first-party sources, draft them, check them against a thirteen-point gate and either publish or hold them. Held pages are reworked rather than discarded. The current state of all of that — corpus size, hold queue, source health, last run of each routine — is published on the health page, because a system that fails silently is worse than no system.