How the weekly “AI in mental health” brief is built

When: Mondays at 14:00 UTC (about 9 AM Central), as a GitHub Action, after the week's daily digests.
What it does: gathers the domains of recent digest survivors with their counts, top relevance and a sample headline, and asks the model which look like content farms, press-release republishers or fake local news.
Output: a pull request adding the flagged domains to the digest's block list. It never merges itself ; Suman reviews it.

An interactive map of a two-part system: a daily news digest that casts a wide net, and a weekly pipeline that re-verifies everything, writes the brief, and renders the edition readers open. A publishing step then puts the edition online at mentalhealthnews.ai and emails subscribers, and a set of scheduled checks watches the whole thing.

Click any box to expand it for the exact sources, search terms, and rules. Use the controls below to expand everything or to highlight where an AI model, rather than a fixed rule, does the judging. Every AI box names the model it runs on today.

Category guide: how stories are sorted