The complete retained text of every published signal's underlying article: curated source text where it exists, otherwise the page text the discovery pipeline fetched.
One row per published signal that has text; curated source text wins over pipeline-cached text. This is a working corpus for internal research, provenance, and quotation. It is not a redistribution channel: do not republish articles outside the company, and link to the original source when sharing outward. Downloading requires an access key.
| Column | Type | Definition |
|---|---|---|
| signal_id | text | The published signal this text belongs to. |
| signal_title | text | Title of the signal. |
| published_on | date | Signal's editorial date (YYYY-MM-DD). |
| significance | enum | high, medium, or low. |
| lenses | text | Audience lenses, joined with a semicolon. |
| source_title | text | Title of the underlying article. |
| outlet | text | Publishing outlet, when known. |
| source_url | text | Link to the original article. |
| text_chars | number | Character count of the retained text. |
| full_text | longtext | The complete retained article text. |
unfiltered
Heavy dataset: the preview is capped at 25 rows; download for the full corpus.
Preview needs an access key. Unlock once and it opens up here.