Posts on GitHub Highlights are drafted with a language model from public repository data, and every one ends with the sources it was written from, each linked and dated.
How a project gets picked
A script queries the GitHub search API for repositories created or revived recently, keeps a local record of star counts so it can compute what changed since the last run, and drops forks, archived repositories, anything without a licence or a real README, and the list-and-tutorial genre. What survives is ranked and handed to a writer as a brief. No project can pay to be covered and none has.
What the model does and does not do
It works from the repository metadata, the README, release notes and linked discussion. It does not invent numbers, dates or quotes. Every figure in a post comes from the snapshot recorded in the post's own front matter, taken on the date shown beside it.
Why the numbers have a date on them
Star counts, release tags and last-push dates are captured once, when the post is written, and never re-fetched. A number that silently updates would make the surrounding sentence wrong without anyone noticing. A dated snapshot is allowed to age.
Review before publishing
A person reads every post before it goes live and checks that the sources exist and that the post reflects them. That is a fact check, not a full copy edit.
Mistakes
A post found to be wrong is corrected, and a dated line at the bottom says what changed. Nothing is deleted quietly. See corrections.
Copyright and trademarks
Posts are summaries in our own words, not translations or copies. Direct quotes are short and attributed. Code snippets are the project's own install or usage commands, used as documentation and covered by the project's licence, which each post names. Project names and logos belong to their owners. This site is not affiliated with GitHub, Inc.