Datasets
Everything the desk computes, it also publishes as data. This is the whole open-data surface in one place: the same records the pages show, in machine-readable form, free to reuse with attribution. The /node protocol list says the same thing to machines; this is its human-facing twin.
every multi-piece story chain the desk tracks: members in order, status, and what would reopen each.
Data: /chains.json · CC BY 4.0
The desk's flagship finding as a standing dataset: across every cross-outlet audit, what share of the split is framing rather than fact, month by month, with per-outlet appearances.
Data: /framing-index.json · CC BY 4.0
The flag index: every word the desk has flagged on its own face — euphemism, characterization worn as fact, a state left ambiguous — grouped by kind, each span verbatim from the linked piece.
Data: /flags.json · CC BY 4.0
Every cross-outlet story the desk framed apart, with each outlet's verbatim angle laid side by side — how N outlets framed one event, gathered across every coverage brief.
Data: /angles.json · CC BY 4.0
A dictionary of contested language: every word the press has used to mean two things at once, organized by the word (never by outlet), with every reading verbatim.
Data: /lexicon.json · CC BY 4.0
A preview of the deterministic checklist layer: five rule-detector language features counted per 1,000 words and pooled at outlet-type level, with the receipts. Rates only, no ranking.
Data: /rhetoric.json · CC BY 4.0
Who pays for the news: the ad breaks that aired on CNN, MS NOW, and Fox News in one capture window — per-channel ad-load, category mix, and top advertisers, with verbatim ad-copy receipts. Counts and shares from a single window, never a verdict on the networks.
Data: /sponsors.json · CC BY 4.0
Grep the networks: every spoken line captured across CNN, MS NOW, and Fox News in one capture window, each with a channel, a UTC timestamp, and a program / ad / promo tag. A machine-readable record of what was broadcast, searchable — not comprehensive, never a verdict.
Data: /transcripts.json · CC BY 4.0
The whole live record as one file: every published piece with its kind, its counts, its source citations, and — for a cross-outlet piece — its framing rate, plus corpus-wide totals.
Data: /corpus.json · CC BY 4.0
The desk's public accountability ledger: every logged error, the page it touched, what was wrong, and what changed — as structured data and as a subscribable feed.
Data: /corrections.json · /corrections/feed.xml · CC BY 4.0