👋 About bignews.day

Our newsman, waving hello

Nobody's immune to a good story

We read the news the way a folklorist reads tales: what happened, and the shapes of the stories people tell about it, left and right alike.

📰 What it is

Every hour we read the front pages of 262 news sites, from far left to far right, and group their headlines into stories: who's running each one, how prominently, in what mood, and for how long. We follow each story on through TV captions, radio newscasts, over a hundred podcasts and shows across the spectrum, what people retell on Bluesky and Mastodon, and what fact-checkers make of it.

🧩 What's different

Underneath the news is a smaller set of stories people keep telling: they're going to steal the election, the law is being used against enemies, the leader is too old. Folklorists have catalogued the shapes of folktales for a century; we do the same for what's said about politics and public life, in the BND index. Each motif names a shape, with the claims told in it, the motifs that answer it, and the shows on the left, center and right it came up on. Often both sides tell the same story about each other: see them across the aisle.

Our newsman, Scoop McBlurb, with his coffee

⚖️ What it isn't

  • Not a fact-checker. A motif describes how a story is told, never whether it's true. Where a fact-checker has looked into a claim, we link to their verdict.
  • Not on a side. We file what tellers say, not why we think a story spreads, and the same shape gets the same name whoever tells it. Claims carry no party labels; the shows and outlets are marked by their lean so you can see who tells what (🫏 leans left, 🐘 leans right: a lean, not a party).
  • Not finished. A language model running on our own computer suggests motifs and files claims; nothing it files is shown until it's been checked. Claude (Anthropic's AI model) does most of that checking now, reading each claim with its sources; the person who runs the site set the rules and spot-checks Claude's calls. Expect some loose fits, and tell us when you find one. How we test the machine is in the Lab.

🔐 What we collect from Bluesky and Mastodon

  • A sample, read for its stories. We regularly listen to Bluesky's public stream for a few minutes at a time, and read a few pages of Mastodon's public timeline from people who chose to be searchable. We look for the stories people retell, counted across many posts.
  • No names. We keep a post's text and time and a salted fingerprint of its author, enough to count how many people said something, never to say who. No handles, images or links' previews. Labeled posts are skipped, a post deleted while we listen is deleted here, and raw posts are dropped after 30 days.
  • Nothing republished. We never post anyone's words as our own. A few posts that were already widely seen are shown embedded from Bluesky itself, so if the author deletes one it disappears here too. Mastodon posts are never shown. Datasets of posts are never published.

🙋 Who makes it

One independent researcher, with a lot of help from open language models running at home. No funding, no ads, no party.

🔔 Notifications

Follow a story or a saga with the 🔔 on its page and we'll tell you when it moves: back on the front pages, spreading past 30 or 50 of them, or a new part. Or have the big ones without following anything: 🚀 XTREME, ⚡ EXTRA, 📺 the evening news and 🗞️ the week's issues. No account, no email.

🛠️ Want the details?