What AI changed today — observed, signed, anchored. Real-time feed across six continuity streams: training-data omissions, axiom ledger, model presence on registries, arXiv retractions, vendor robots.txt changes, and external seals. Every event is reproducible from the underlying JSON.
AI systems mutate continuously. Model cards get edited, training-data disclosures change, vendor robots.txt allows or blocks the same crawler from one day to the next, papers get retracted, claims appear and disappear. The Observatory is the place where each of those events is captured the moment it happens, written to a signed ledger, and made permanently citable.
Each event in the feed comes from a real underlying JSONL stream that you can inspect, replay, or verify cryptographically.
Use @crovia/seal (or pip install crovia-seal) to produce your own continuity receipts;
/registry/ exposes the full historical archive.