Knowledge hub
The part that is meant to still be true next year.
Plain explanations of how AI-era discovery actually works — what the terms mean, how the measurements are taken, and what to do about each one. Written to be linked to, revised in place, and dated so you can tell.
- Format
- Short explainers, worked examples, no gate
- Upkeep
- Revised in place, with the date on the page
- Scope
- How AI-era discovery works, and how to act on it
0 explainers
How the hub is kept
Five rules, applied to every entry
Reference material is only worth linking to if it is still right when someone follows the link. These are the rules that make that likely, published before the entries rather than after.
- 01Answer the question in the first paragraphThe definition goes at the top, before the context and before the argument. Someone who arrived from a search wanted one sentence, and making them scroll for it to raise the time-on-page is a tax charged on a reader for our benefit.
- 02One entry per ideaAn explainer covers a single term or a single task. Two ideas in one page means neither can be linked to precisely, and precise linking is the entire reason a reference layer exists.
- 03Show the version that breaksA how-to that only demonstrates the happy path is not a how-to. Each one states what it assumes, and what happens when that assumption does not hold.
- 04Revise in place, and date the revisionAssistant behaviour changes quarterly, so a page written against last year's behaviour is wrong rather than merely old. Entries are rewritten under the same URL and carry the date they were last checked — a new post correcting an old one leaves the wrong answer in circulation.
- 05Retire what is no longer trueAn entry that cannot be made correct is removed rather than left standing with a caveat at the top. A hedge does not stop a stale page from being the one that ranks.
An entry that cannot meet these is retired rather than caveated, which is why the shelf above is shorter than it could be.