What we read to find the news, and every publisher our briefs have cited.
How it works
Several times a day we read the RSS feeds and X accounts listed below. When something new appears, we look for the primary source: the announcement, paper, filing or post where it started.
A brief needs at least two distinct publishers. When an official announcement is enough on its own, the brief says so and is marked as a single source. We summarize in our own words and link to the originals instead of republishing them, and each brief is reviewed before it goes online.
Read our editorial standards for how we handle evidence, dates and corrections.
By the numbers
Computed from every published brief each time the site is rebuilt.
Briefs published
303
Publication days
292026-10-22 – 2026-11-20
AI / robotics
211 / 9270% / 30%
Briefs citing two or more publishers
29397%
Average importance
3.8out of 5
Publishers cited
389
Publishers cited
389 publishers named as the primary source or as coverage in 303 briefs, most cited first. A publisher counts once per brief. Names are matched ignoring case and spacing; different names are never merged. Where a publisher was cited through another site, such as a newswire or a code host, that site is named and the publisher is not linked.
37 feeds, read on a schedule that depends on how much signal and how much noise each one carries. Research preprints, general-interest sites and aggregators are filtered for AI and robotics before anything is kept. Feeds that currently fail to load are not listed.
25 accounts from labs, companies and research groups, searched in every round. We also run open searches on AI and robotics topics, because important news does not only come from accounts on a list.