A reading room
Research is only as good as the request
that produced it. So the request comes with it.
A publishing surface for public-records work. Every post says which request produced it, which agency answered, under which jurisdiction’s law, on which dates. Everything arrives as a draft, a redaction scan runs before anything is public, and there is no algorithm deciding who reads what.
What this is
A place to publish what a records request actually returned, alongside the request itself. Not a feed, not a newsroom, not a document dump — a reading room, organised by subject, where each piece of research can be checked back to the paperwork that produced it.
Most published records research is an assertion with a document attached. Somebody obtained something, somebody wrote about it, and the reader is asked to take on trust that the document is what the writer says it is and that it came from where they say it came from. That trust is usually warranted. It is also unverifiable, which means it collapses the moment anyone contests it — and records research about powerful institutions gets contested.
So the unit of publication here is not the document. It is the request: an agency, a jurisdiction whose law it was filed under, a date it went out, a date something came back, an outcome, and then the documents and the author’s reading of them. A reader who wants to check the work has the reference they need to file the same request themselves.
The spine · 1
Nothing publishes on arrival
Research reaches the reading room through an API, usually from the records console that filed the request. That call never publishes. It creates a draft, runs a redaction scan across everything submitted, and returns what the scan found. A human opens the draft, reads the findings, and decides.
This is not a workflow preference. Records released under public-records law routinely contain third-party personal information that an agency failed to redact. A one-click path from a document dump to a public post is precisely how a witness’s home address ends up on the internet.
The spine · 2
The scan is a net, not a guarantee
The scan looks for the patterns that matter — names beside dates of birth, telephone numbers, addresses, identification numbers, medical language, anything suggesting a minor. Findings block the publish action until someone resolves each one.
Two categories do not route to the author at all. Anything reading as a minor or as medical information goes to review, because those are the two where an author’s own judgement under deadline is least reliable and the harm is least reversible.
And the scan misses things. It is a net. Privacy and redaction →
It is a reading room, not a feed
There is no infinite scroll. There is no follower count on a post, no view count, no like, no trending list, and nothing that ranks one author’s work above another’s because it moved faster. Following an author or a subject produces a list you asked for, in the order things happened.
This is a design decision with a specific reason behind it. The moment a number on a post can be moved by publishing more, publishing sooner, or publishing something more alarming, the incentive stops pointing at accuracy. Records research is unusually easy to farm this way: a document looks like evidence whatever it actually says, and the most inflammatory page of a 400-page release is the one that travels. A reading room that amplified by engagement would reliably surface the least careful reading of the most sensational document.
Who is publishing
Pseudonymity is allowed, and normal.
Some of this research is about people with lawyers, budgets, and long memories. An author may publish under a persistent pseudonym; what has to be verifiable is the request, not the requester.
A pseudonymous author is not an anonymous one. A handle is persistent, its posts accumulate under it, and its record of corrections and takedowns is as visible as anyone else’s. What is not required is a legal name attached to work about a sheriff’s office in the county where the author lives.
This costs something and it is worth saying what: a reader cannot weigh an author’s institutional standing if there is no institution named. The answer is that the provenance line does that work instead. The agency, the jurisdiction, the request identifier and the dates are checkable by a stranger without knowing who filed it.
What a pseudonym does not buy
It is not a shield against the platform’s own rules, and it is not a shield against a court. Takedown applies identically. And publishing a document does not become safer because the byline is a handle — the responsibility for what is in the document is the author’s either way.
Your research, on your own domain
One script tag renders an author’s published research on their own site, so readers land there rather than here. It runs in a shadow root, sends nothing about the reader, and if this service is unreachable it renders nothing at all rather than a broken box.
What is on this site
- How publishing works — draft, scan, review, publish, correct. The stages, and the one that cannot be skipped.
- Provenance — the six things every post carries, why a missing one is a hole and not a detail, and what provenance still does not prove.
- Privacy and redaction — what the scan looks for, what blocks, what routes to review rather than to the author, and where the net has holes.
- Topics — the shared vocabulary: eighteen topics, fifty-five subjects, and the record types that actually exist for each.
- The widget — both directions, every option, and a plain statement of what it sends.
- Following and collections — what a follow does, and every metric that deliberately does not exist.
- Takedown — how to ask, what happens, and the deadline this service holds itself to.
- Limits — four things this platform does not do, stated as clearly as the things it does.
A published document is a document that was published.
It is not proof that the document is genuine, complete, or means what the person publishing it says it means. This platform does not verify any of those things, and a reading room that let you believe otherwise would be worse than no reading room at all.
How this works
0:56 · narrated · captioned · served from this site, and nothing is sent anywhere when you press play. Prefer to read it? The transcript is below.
Transcript
| At | Section | What it says |
|---|---|---|
| 0:02 | What it is | A reading room for public-records research. Every post is published with the request that produced it, so a reader can go and check the work themselves. |
| 0:11 | Provenance | Which agency answered, under which jurisdiction's law, filed on which date, answered on which date, with what outcome. Research without that is an assertion. |
| 0:21 | Draft first | Research arrives over an API, and that call never publishes. It creates a draft, runs a redaction scan, and hands the findings back before anything is public. |
| 0:32 | Not a feed | No infinite scroll, no view counts, no trending, no algorithm. A number on a post that can be moved by publishing more stops pointing at accuracy. |
| 0:43 | The limit | Publishing a document does not make it true, and this platform does not verify that a document is what its publisher says it is. It makes the claim checkable. |