Everything public about the organization the Atlas is written for, since ChatGPT: its SEC filings and press releases, news, research papers, AI patents, comment letters and regulatory documents, plus the AI timeline and the cited profile built from them, in one row shape.
Built once on 2026-09-27 by a local backfill (docs/self-record.md) and refreshed when it is rerun. One row per document (record_type document), per AI timeline event (timeline_event) and per sentence of the cited profile (profile_sentence). Documents carry the AI paragraphs found in them and a text excerpt; events and profile sentences carry the ids of the documents they cite (cites) and those documents' URLs (cite_urls). Public sources only; every event and sentence traces to a document row. With ?company=<slug> the same record is served for another tracked company, where one has been built. The engine's ongoing news, facts and metrics for the same organization are in intel-items, intel-facts and intel-metrics with ?company=<slug>. This is a working corpus, not a redistribution channel. Downloading requires an access key.
| Column | Type | Definition |
|---|---|---|
| record_type | enum | document, timeline_event, or profile_sentence. |
| record_id | text | Stable id: a document or event UUID, or <slug>:profile:<n> for a profile sentence. |
| company_slug | text | The organization's registry slug; joins intel-companies. |
| company_name | text | The organization's name, from the registry. |
| date | date | Publication date for a document, event date for an event, build date for a profile sentence (YYYY-MM-DD). |
| source | text | For a document: sec_filing, sec_exhibit, news, newsroom, paper, patent, enforcement, comment_letter, merger, or testimony. For an event: its category. For a profile sentence: profile. |
| title | text | Document title, event headline, or "Profile sentence <n>". |
| text | longtext | A document's one- or two-sentence factual summary, an event's body, or the profile sentence. |
| url | text | The public document's URL (documents only). |
| ai_related | enum | yes when the row concerns AI, machine learning or data science at the organization. |
| dimension | text | For a document: the Intel Desk dimension it was filed under. |
| ai_passages | longtext | For a document: the paragraphs that name AI, separated by blank lines. |
| text_excerpt | longtext | For a document: the start of its text (up to 8,000 characters), when retained. |
| cites | text | For an event or profile sentence: the record_ids of the documents it cites, separated by semicolons. |
| cite_urls | text | The URLs of those documents, separated by spaces. |
| metadata | text | For a document: source-specific fields as JSON (form and accession, patent number and CPC codes, DOI and venue, agency and docket). |
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.