# AsOf Data

Open snapshots with a DOI.

Periodic dumps of the full ledger: entities, claims, sources and the links between them, as JSONL and Parquet, published on Hugging Face with a Zenodo DOI so papers and benchmarks can cite a fixed version.

## Latest snapshot

Snapshots are produced monthly by pipeline/snapshot.py. Until a Hugging Face dataset is configured, the live API, /llms-full.txt and the markdown twins are the record.

## Sources

When each source last put a new claim on the record. Structured sources refresh daily (Wikidata on Sundays); feeds every weekday.

| source | last claim | age |
|---|---|---|
| wikidata | 2026-10-27 | today |
| wikidata-products | 2026-10-27 | today |
| sec | 2026-10-28 | today |
| feed | 2026-10-28 | today |
| huggingface | 2026-10-27 | today |
| google-status | 2026-10-27 | today |
| verification | 2026-10-28 | today |
| manual | 2026-10-27 | today |

## Integrity

Measured from the live database, not asserted. Hard rules must be zero; the same numbers are at /v1/health.

| measure | value |
|---|---|
| claims · entities · sources | 56,121 · 477 · 19,308 |
| claims with a real date | 89.4% |
| claims with a primary source | 80.7% |
| valid · unverified · superseded · flagged | 10,937 · 10,597 · 34,585 · 2 |
| unsourced claims (hard rule) | 0 |
| dangling supersession pointers (hard rule) | 0 |
| claims ending before they start (hard rule) | 0 |
| future-dated claims (hard rule) | 0 |
| open duplicate groups (hard rule) | 0 |
| values still unlabeled | 3 |
| latest change on record | 2026-10-28 |

## License

Claims are facts and carry no copyright. Descriptive text is CC BY 4.0. Text derived from Wikipedia is CC BY-SA and attributed. Wikidata identifiers are CC0. SEC filings are public domain. Attribute AsOf when you publish.