Federal Hill
Bardstown, KY — 1922–present
A fabricated slave cemetery presented as history for over a century, despite archaeological proof it never existed.
Investigative journalism · Louisville, KY
Every file follows the same path: the public claim, the record underneath it, and the property that connects them. A file marked RESEARCH carries a question, not a conclusion. Nothing here has published yet.
Bardstown, KY — 1922–present
A fabricated slave cemetery presented as history for over a century, despite archaeological proof it never existed.
What’s open right now. Records requested, records produced, and the documents already scheduled to arrive. No findings. This file hasn’t published.
Louisville, KY · opened August 2026
A 153-acre data center approved beside Rubbertown, the historically industrial, majority-Black stretch of southwest Louisville. It cleared after a moratorium sat in committee for eleven months and expired unpassed. A resident made six specific claims about it in public comment at Metro Council on 13 August 2026. The file checks all six against the record, one at a time.
Held in the file
The Kentucky Department of Revenue reported on 4 May 2026 that it had received no notification of any approved data center project in the Commonwealth and had issued no certificates of exemption. That’s the department’s account of its own program, and it’s four months old.
Already scheduled to arrive
Audio and documentary work. Nothing has aired yet. What follows is what’s being built and the rules it’s being built to. It isn’t a schedule, and there’s no date attached to it.
One case per season
A season follows one case from the first records pull to the final draft, including the parts that collapse. The killed-speculation log is part of the material, not an appendix.
Field and document
Built on the same five-phase method as the written work. Every claim sourced on screen, every document shown unaltered. No straightening, no whitening, no scan used as a background texture.
Testing claims, one at a time
The first series in production takes six claims from a single public comment and checks each one against the record. It’s the format where the standards are easiest to break, which is why it’s the one being built first.
No air date, because there isn’t one yet. When something ships it shows up here with the documents attached.
What’s actually published, and where to read it.
Diaspora. Digest
Stories from the global Black world.
Bubba · Louisville, KY
A weekly read on the Black diaspora. One big story, five things from five places, and an Underreported section for what got slept on. At least three of the five have to be non-U.S. That rule is what keeps it a diaspora letter instead of a Black-news roundup. It’s separate from the investigations: different audience, different voice, same ledger discipline with a lighter touch.
Five issues published, 10 May to 8 June 2026. It’s between runs right now and coming back; the sourcing pipeline restarted on 1 September 2026.
Version 0.1 · Zenodo, 25 May 2026 · CC BY 4.0
The research methodology, written as a specification rather than a manual: how a claim is tiered, how a source is classed, and what a case isn’t allowed to say. It’s built for cases where the investigative file is itself compromised, managed or incomplete. Treat the official record as a neutral starting point in one of those and you end up working inside a narrative somebody else built. Version 0.2 is in revision against seven cases and adds Records Debt as a named component.
Version 0.1 · drafted 5 September 2026
How research becomes a published claim: the five claim tiers, the six source classes, how a quotation is handled, what counts as ours and what counts as the record, and what happens when we get something wrong. The Standards section of this site summarizes it. The full document publishes here.
Nothing published yet
Written investigations are drafted for outside publication, and peer-reviewed contributions are planned. Neither has run yet. When one does it gets listed here with its outlet and date, and not before.
DBS doesn’t generate truth. The cases are already there, in archives, court records, the ground, and people’s memories. So almost everything that can go wrong here is a handling problem. These rules make a handling failure visible instead of invisible, and every one of them says what happens when it can’t be met.
Every factual claim carries a tier before it is written down, assigned in the research file rather than at the draft stage.
Below all five sits the audit gap: something we believe, paired with the exact record that would settle it, and nothing else. It isn’t a finding and it isn’t publishable in any form. Three open cases have an empty Documented tier by design, because they were opened from working knowledge instead of from records. Their first milestone is one documented claim, and it doesn’t exist yet. Availability is not custody. A footnote is a map to a record, not the record.
A managed source is attributed out loud
A record produced by an institution that’s a subject of the investigation can’t be published as fact on its own. Either it’s named in the sentence that uses it, or it’s corroborated independently, or it doesn’t run. Institutions get split class by question, not by name.
Never a card without a document behind it
A ledger card quotes one verbatim line from a primary source and is followed immediately by the unaltered scan. The card heralds evidence; it never substitutes for it. Once per piece, twice if the story genuinely has two.
Documents keep their paper
Skew, stains, fold lines and the crop as it came off the glass. Nothing is straightened, whitened, sharpened or recropped, and a scan never appears as texture behind a title. The moment a real document becomes decoration, you can no longer tell which part of the frame is evidence.
Quotations are not cut inside a sentence
Original spelling, capitalization and errors are preserved exactly. A misspelled name in a 1965 police report is information. Recorded speech by a living, named person is never cut inside a sentence, in any medium, for any reason.
Killed angles stay killed, and stay visible
When an angle collapses it gets logged with the date, the reason, and what survives the kill. It isn’t revived without new documents, and never in a softer form: not hedged, not as a question, not as "some have wondered."
The furniture never touches a case
Period broadcast styling carries the show open, the transitions and the lower thirds. It doesn’t carry a claim. When the two conflict, the styling is the negotiable half.
Corrections are their own item
Same length, same places as the thing they correct. The original stays up. A correction states what was wrong, what is right, the date, and what else the error touched. Claims keep their identifier when corrected, struck through with a dated note, so an old citation still resolves.
Some documented material stays out
Private relationships that are not the engine of what happened. Vivid, sourced, irrelevant detail. People investigated and never charged. No true-crime register, no "victim’s final hours." The people in these cases get the dignity of the record.
No corrections have been issued, because nothing has published yet. When one is, it runs here at full length next to the piece it corrects. And if you find a document on this site that’s been cleaned up, straightened, or used as a background texture, that’s a violation of the standard and worth writing to us about.
DBS is an investigative journalism operation built on a single premise: property is the motive, the deed chain is the story, and the institution is the cover. Every investigation follows the same five-phase methodology: tracing public claims back through their citation chains, following the money and the land through property records, mapping how misinformation propagates through the information ecosystem, pressure-testing every conclusion against alternative explanations, and structuring every publication so the audience can verify it independently.
Solo operation. No newsroom. No editorial board. One researcher, public records, and the question nobody asked.
Every investigation runs on public records requests, archival research, deed searches, and FOIA filings. No advertisers. No sponsors. No editorial interference. The work is funded by people who believe the public record should be public.
No advertisers. No sponsors. No strings.
Fund the records