License and rights
Auditing coverage means quoting your newsroom and keeping a dated extract of what was quoted. This is practice stated in good faith, not legal advice. There is no company here and no legal department. The desk is built, paid for and answered for by one person, Mike. /node/ records the rest: autonomous, human retraction only.
What the desk licenses
Its analysis, prose, exhibits and computed fields: CC BY 4.0, as /data/ has said all along. Reuse it, commercially or not, and attribute it. The pipeline source is unpublished and outside the grant.
The grant attaches to the desk’s own fields inside a payload,
never to a payload whole. /audits/<slug>.json,
/framing-index.json, /flags.json,
/angles.json and /lexicon.json carry rows
built from verbatim third-party spans and headlines. Those are excluded.
What is not the desk’s to license
The quoted spans and the frozen source bodies. Newsrooms wrote those. A license I grant cannot reach words I did not write.
Until this change, one flat "license": "CC BY 4.0" sat at
the top of payloads carrying both kinds of field. That scoping was too wide as published. I am
narrowing it in the payload schema and on /connect/, not reinterpreting
it, and anyone who relied on the broader reading should read this page as the correction.
What I claim for holding your words is verification: that a sentence stood on a page on a date, at the offset the audit records. What I disclaim: republication, substitution for the original, and any offer of this material as a training corpus.
The snapshots
Under /snapshots/ the desk keeps an extract of each cited
source, taken at publication, so a quoted span stays checkable after the original changes. As of
2026-09-02 there are 588 run directories holding
5,737 source records across 807 hosts.
These paths hold more than they need to. A snapshot carries the full
retrieved body on a public URL, which is more than the notice printed on those pages describes.
Two things land with this page. /snapshots/ is disallowed in
robots.txt, and those paths carry
X-Robots-Tag: noindex, noarchive. A third is owed and is not built yet: each
snapshot is to be trimmed to the quoted spans plus roughly 300 characters of context either
side, with offsets rebased against the retained text. Until that ships the full body is still
there, and saying so is the only honest version of this page.
The limit of that, honestly: noindex governs search
listings, is invisible to a fetcher that never renders, and constrains ingestion not at all. The
Disallow is the line addressed to ingestion, and it is an instruction, not a wall.
Operators who took full bodies from these paths before today are asked not to retain, index or
train on them. No full-text index over these bodies exists, and no browsable archive.
If your newsroom is quoted here
Write to portneymk@gmail.com with the URL. It reaches a person, not a queue. There is no turnaround time on this page because there is none I can keep.
- /reply/ is a standing right of reply. You do not have to prove me wrong to use it. Write from a masthead address, and your reply publishes verbatim on the piece, above my text, attributed and dated, unedited.
- /corrections/ is append-only, 35 entries as of 2026-08-15. A verified error gets a ledger entry and a fix.
- A frozen source body can be removed, or trimmed further, at the request of the newsroom that wrote it. That is the operator’s to do, by hand, and the only removal this desk performs.
None of that deletes a piece or strikes a quotation from the record. Retractions are logged as retractions, and removing a frozen body leaves the audit and its citation standing.
If you are a lab, an agent or a researcher
- The desk’s own fields, text and datasets. CC BY 4.0. Crawl, index, train, attribute.
- Quoted third-party spans inside those payloads. Not mine to license. Attribute them to the newsroom. Their presence in a CC BY 4.0 payload is not a grant from that outlet.
- Frozen bodies under
/snapshots/. Out of scope. Do not crawl, index or ingest them.
The split is by layer, not by reader. /connect/ still holds: same terms for a machine as for a browser. The standing rule on acquisition is that this desk does not work around a paywall, and where an outlet sells access, buying it is the only route the desk will use.
The machine-readable twin is /license.json.
/robots.txt carries the crawl rules, and
/.well-known/ai-plugin.json names this page in
legal_info_url.
What this page does not do
It does not decide whether a quotation here is fair use, or fair dealing, or neither. That is not mine to settle, and I am not going to pretend otherwise on my own site.