agitracker.io

Sources and method

Where the AI news feed comes from, and exactly what it does with it.

How it works

  1. Collect. Once a day, each source’s public RSS or Atom feed is requested once, with a descriptive user agent.
  2. Filter. Feeds that cover more than AI (general tech news, company blogs, press-release wires) are filtered by AI terms; wires must mention AI in the headline.
  3. Store only the minimum. Headline, link, time, source and an excerpt of at most 220 characters taken from the feed. No article bodies and no images are copied. Every headline links to the original.
  4. De-duplicate and group. Tracking parameters are removed from links, and near-identical headlines from different outlets are grouped into one story.
  5. Rank. In the weekly brief, a story scores higher when it comes from a lab or company directly, and when more independent outlets reported it. Press releases are listed separately.
  6. Label. Topics are assigned by keyword rules. They are a convenience and will sometimes be wrong.

Links to press-release wires are marked nofollow because those releases are written and paid for by the issuing company. Inclusion of a source or story is not an endorsement. 649 headlines are currently held in the rolling window; completed weeks are archived as briefs.

AI labs 8

Companys 9

Publications 17

Researchs 2

Newsletter / communitys 3

Newswires 2

Not covered yet

Some organisations publish no public RSS feed (several frontier labs among them), and some newswires block automated clients. They are skipped rather than scraped, so their announcements appear here only when another source in the list reports them.