A lot happens in AI every day. A few things are worth your time.
OpenBeat reads 17 publishers’ own feeds, groups the reports of each story, and turns the strongest into short English posts. An AI desk does the reading and drafting; a Fact Checker checks every claim; a human editor approves every post. Free, no sign-up to read.
How a story gets here
Live numbers from our database. The marked steps are where people, or a separate checker, stand between the AI and you.
- 01CollectAround the clock
Publishers’ own feeds, release notes, repositories and channels on the watchlist.
17 feeds - 02GroupOne story, one place
Reports of the same story are folded together; more independent sources mean more heat.
183 reports → 177 stories - 03ScoreLess is more
A small model scores each story; company sources need 60, media 76. The watchlist can raise it.
5 drafted - 04WriteFrom the original
The Writer reads the original page and drafts in plain English, with a code example or our own chart when the source has one.
- 05CheckEvery claim
A separate Fact Checker compares each claim with the source and can only send the draft back.
0 sent backHuman or checked step - 06ApproveA person decides
An editor approves or rejects every post. Nothing is published automatically.
5 publishedHuman or checked step - 07CorrectIn the open
Readers report errors; corrections are shown on the post and listed publicly.
0 correctionsHuman or checked step - 08DeliverOne source, many outputs
The site, Picks, Hot, daily, weekly and monthly issues, RSS, Atom, JSON Feed and IndexNow.
1 days of posts
What else you’ll find
OpenBeat summarises and links to the original reports; their rights stay with the publishers. To ask for a correction or removal, use “Report an error” on the post. Parts of our news radar are based on the open-source AIHOT engine (MIT).