THE FEED
Read everything. Trust what's checked. See the whole picture.
One free daily paper, built on sources you can see and a model you can inspect. It shows what happened, not what a feed decided you should see.
What you get
An editor that shows its working.
One story, many outlets
A story carried by ten outlets is one story with ten sources, not ten cards. Ranking weighs impact and breadth, and old news ages out.
Briefings you can check
A daily briefing, cited and verified against its sources. On a quiet day it says so instead of padding the page.
Perspectives
The biggest stories read across every outlet that carried them: common ground, framing and what is in dispute. It describes framing. It never rules on who is right.
The scorecard
Sourcing, manipulation signals, outlet record and corroboration, each shown as its own number beside the story.
The Test
New. For the day's top stories: the claims, the best case on each side, what would settle it and what is still unanswered. It ends on questions.
Memory
Pattern Watch follows the week's through-lines. Month in Review rewrites the last 30 days every Sunday.
The scorecard
Not a verdict. The parts.
A single trust number asks you to take its word for it. The Feed breaks a story into the things you would check yourself, and a story nobody has scored says unrated instead of guessing.
What a scored story shows
- Scrutiny
- How much care to take. Higher means more care.
- Sourcing
- How well the piece backs up what it says.
- Manipulation
- Fallacies, propaganda techniques, slop and loaded language.
- Outlet
- Wire service, established outlet, or blog and unknown.
- Corroboration
- How many other outlets carried it.
The algorithm is printed
One run a day, in this order.
Nothing is tuned to you, and the main ranking runs on reach and coverage, not clicks. It all happens once a day on local hardware, with few enough model calls to count.
Ingest
Fetch every source, remove duplicates, group stories.
No modelImpact
Rate the day's 30 biggest stories by reach.
ModelBriefings
Write the desk briefings and Pattern Watch.
ModelPerspectives
Read the top stories across outlets, and put five to The Test.
ModelScore
A small daily quota of full scorecards, never bulk.
Model
What it will not do
Follow you around
No analytics and no third-party scripts. Images are fetched ahead of time and served from here, so publishers never see your browser.
Ask you to sign up
All nine desks are open. Accounts exist only for extras, such as saved stories and your own model API.
Make things up
On a slow news day it says so. If no story was carried by at least two outlets, the briefing reads “A quiet day. Nothing moved that changes the picture.” instead of filler. And if the model server is down, the news still shows and no fake score is written.
Today's paper is already out.