Our process
How we build
Short answer: pages on this site follow the same release checklist, and a person approves a release before it is published. This page describes our process. It is not a promise of results.
What this means for you
- 1Planned is labelled plannedWe label things that are not open yet as planned, and nothing is sold before it opens.
- 2Notes carry a status tagStudy notes say where they come from and whether a point is verified or inferred.
- 3A person signs offA release is published only after a person approves it.
The release checklist
- 1Input checksWhere a page or tool takes input, that input is checked before it is used.
- 2Sign-in on private areasSome areas, such as access to a purchased library, may use sign-in or a browser cookie. Any such area is planned to sit behind that check before it opens.
- 3Written docsWe aim to give each tool and study note a plain page that says what it does and what it does not do.
- 4Tests before releaseBefore a release, automated checks run across the site's pages: wording rules, broken links and images, page errors, brand colours and source status tags.
- 5Rate limits where they applyWhere this site has a service you can call, such as purchased-library access, request rates on it are limited or are planned to be limited before it opens. This is not a guarantee of security.
- 6A working public addressEach release is published to the site's address; the previous version is kept so it can be restored.
A person approves the release
- 1DraftPeople and AI tools draft in a staging copy, not on the public site.
- 2ChecksThe automated checks above run on the staging copy. A failed check keeps it unpublished.
- 3ReviewWe review the wording and run security checks; this is not a guarantee of security.
- 4ApprovalA person approves the exact release. Nothing is published automatically.
- 5After releaseWe look at the result using only our own records, and restore the previous version if something is wrong.