What we publish
Explainers: pieces that answer a question a person actually asked. We do not publish news, opinion pieces, sponsored articles, or anything written to fill a gap in a keyword list.
A topic qualifies if it has a settled, checkable answer, or if the state of the disagreement is itself worth explaining. A topic does not qualify if answering it responsibly would require expertise we do not have. Medical, legal and financial questions are treated as explanations of concepts, never as advice for your situation.
How this launch set was written, plainly
The explainers published at launch were drafted with substantial assistance from a large language model, working from its general knowledge, and edited for structure and clarity. They were not individually verified against primary sources before publication.
We are saying that here rather than letting you assume otherwise, because the alternative is a claim we cannot currently support. What follows is the standard we hold, and the honest gap between it and where this archive stands today.
Treat these explainers as a well-informed starting point rather than a citable source. For anything that matters, check the primary source before relying on it.
The sourcing standard
The standard is that claims are checked against primary sources where a primary source exists: the specification, the paper, the statute, the dataset, the original text. Secondary summaries are for finding primary sources, not for replacing them. Where a number appears it should come from a named body and carry its date, because a figure without a date is a figure that will quietly become wrong.
Bringing the launch archive up to that standard is work in progress. Pieces are being reviewed in order of how much a reader would rely on them: the ones touching health, money and safety first. Anything found to be wrong is corrected under the process below rather than quietly amended.
On using AI at all
We think a language model is a legitimate drafting tool, in the way a spellchecker is a legitimate tool, and we think readers are entitled to know when one was used rather than discover it.
What a model cannot do is verify itself. It produces text that is plausible, and a plausible wrong answer is indistinguishable from a plausible right one from the inside. That is precisely why the verification step is a separate human obligation and why its absence is disclosed above instead of being papered over.
Responsibility for every published sentence sits with the publication, not with the tool.
Corrections
When we get something wrong we fix it, and we say what changed. Corrections that alter the substance of a piece are noted on the piece itself with the date. Typos and broken links are fixed silently.
To report an error, use the contact form and choose "A correction". Point at the sentence if you can. You do not need to be polite about it, and given what is disclosed above, corrections are especially welcome right now.
Updating
Explainers about moving subjects are revisited. Where a piece depends on figures or milestones that change, it is written so the changing part sits in one section and can be updated without rewriting the whole article. The publication date and the last update are both shown.
Independence and money
Nothing on this site is paid placement. No company pays to appear in an explainer, to be described in a particular way, or to keep something out of one. If that ever changes, it will be disclosed on this page before it appears anywhere else on the site.
Links
Internal links exist to take a reader to the next thing worth understanding, not to move ranking signals around. External links point at sources, and we do not link out as a favour.
Republishing
Quoting a paragraph with a link back is always fine and never needs asking. Republishing a whole explainer needs permission, which is usually granted. Ask through the contact form.