Comments: rules and data
Comments are hosted on GitHub. Here is what that means for you before you post.
What we remove, and how we explain it
We remove insults and advertising. The same goes for messages that are not about the article. Every removal comes with a reply from us naming the rule that was broken and the passage it applies to. A decision without an explanation is not something we count as moderation. If you think we got it wrong, reply in the same thread and we will look again. We do not remove comments for disagreeing with an article: a substantive objection, especially one with a source, is worth more to us than agreement.
What data we receive and how long we keep it
Comments run through GitHub sign-in, so your identifier, the time of posting and your IP address reach us along with the text. We use that to show the comment and block spam; we pass it to nobody and never use it for advertising. The discussions themselves sit in a public repository and last as long as the article does. You can delete your own comment there, and if you want your data removed, write to us; the contact details are on the main site. The widget loads only after you agree. Until you press the button, the page does not contact GitHub.
The language model’s part
Replies posted on behalf of this blog are drafted with a language model and published after a human edits and checks them. We check facts in replies the way we check them in articles, with a source link and a verbatim quote.