AI-MINISTRIES/us-federal-public-data-mirror

US Federal Public Data Mirror

An independent copy of United States federal public datasets that have been removed, discontinued, degraded, or formally proposed for elimination.

Everything here was already public and is United States Government work, which is not subject to copyright under 17 U.S.C. § 105.

Nothing in this mirror is modified. No values are corrected, no records are dropped, no columns are renamed, no formats are converted. Each file is stored exactly as the agency published it, alongside the date it was retrieved and a SHA-256 checksum, so that any copy can be verified against any other copy.

This is not affiliated with, endorsed by, or connected to NOAA, the CDC, the EPA, or any other agency or organisation.

Why this exists

Two different things have been happening to US federal data, and they need different responses.

Publication removals. In January and February 2025, roughly 8,000 pages and about 3,000 datasets were removed from or altered on federal websites. Most of that was reversed: a federal court ordered restoration in Doctors for America v. OPM/HHS/CDC/FDA, and the Harvard Library Innovation Lab independently archived the full data.gov catalogue — over 311,000 datasets, updated daily. That problem is substantially handled by institutions better resourced than this one, and this mirror does not attempt to duplicate it.

Collection stoppages. These are permanent and no archive can fix them. Weather balloon launches were cut or halved at eleven National Weather Service offices in March 2025. Nine Aleutian seismic stations feeding the tsunami warning system went dark in November 2025 over a $300,000 grant. CPI data was not collected at all during the October–November 2025 shutdown, leaving permanent holes in the inflation record. USDA discontinued the Mink Survey and the Agricultural Labor Survey outright. Data that was never collected cannot be rescued.

This mirror sits between those two cases. It holds a deliberately small set of datasets that are small enough to keep current, specific enough to be verified, and either already abandoned or formally proposed for elimination — and it keeps them in a form an ordinary person can actually download and read, rather than only as part of a multi-terabyte bulk research archive.

For several of these, the value is precisely that this is a copy taken before any further revision. A baseline captured now is what a claim about change can be measured against later.

What is here

Exact source URLs, retrieval dates, byte counts and checksums for every individual file are in MANIFEST.json.

What is deliberately not here

• —Anything already robustly mirrored by a major institution. The full data.gov catalogue is Harvard's; duplicating it would be vanity.
• —Query-only sources. CDC WONDER, USGS NWIS and similar have no bulk export and would need purpose-built extractors. They are listed as outstanding rather than quietly omitted.
• —Terabyte-scale raster and satellite products. LANDFIRE, MTBS and similar are beyond what a single small operator can responsibly hold.
• —Anything whose source URL has not been confirmed with a real request. A broken entry that looks fetchable is worse than an absent one, because it reads as covered when it is not.

Verifying a copy

sha256sum -c SHA256SUMS

MANIFEST.json additionally records, per file, the URL it came from, the byte count, and the UTC timestamp of retrieval.

Freshness

Each dataset carries an expected update cadence. A live view showing the last successful retrieval per dataset — and marking any dataset that has gone past its cadence as stale — is at stillhere.ai-ministries.com.

A mirror that has silently stopped updating is worse than no mirror, because someone trusts it. Staleness here is always stated, never inferred.

Reproducing this yourself

The fetcher is a single Python file using only the standard library. There is nothing to install and no account required:

python3 fetch.py

Copies are the entire point. If this mirror disappears, that should be a non-event.

Provenance and honesty notes

• —Retrieval is by conditional HTTP request, so unchanged files are not re-downloaded and their original checksums carry forward unchanged.
• —Some files are legitimately header-only — for example, Storm Events location files for years in which NOAA holds no location records. These are empty tables, not failed downloads, and are kept as published.
• —Files that could not be retrieved are recorded with their HTTP status rather than omitted, so a gap is visible as a gap.