The answer surfaces, as plain CSV files anyone — or any AI — can download. Source-backed, public-domain and open-data only. Every number real, every gap named.
Every plan, issuer, metal tier, structured to compare.
Star ratings + outcome measures. Compare, never diagnose.
Storm + wildfire + FEMA NRI composite, dominant hazard, EAL.
Denial rates by race + Black/White disparity, by state.
Every FDIC-insured institution, active + historical, flagged — verify a lender is real.
LSC-funded legal-aid providers, the free-path referral surface.
The full GLEIF golden copy — every legal entity, for counterparty due-diligence (open data).
Every occupation × industry × wage from the federal record.
Federal-contractor profiles from the public spending record.
Federal buying-office profiles from public procurement data.
These are the source-backed public surfaces. Any caps you see are named, not hidden — tagged honestly, never inflated, and lifted as the larger runs land (the lender directory and provider/entity registries are now the full federal record). Licensed-source surfaces (e.g. the Lightcast skill taxonomy) are kept out of the public files on purpose. Generated from the live unified catalog; refreshed by re-running the export.