How This Publication Works

Every edition begins with a single automated sweep of roughly two dozen RSS feeds across left, independent, mainstream, and conservative outlets, run over a rolling 36-hour window. That produces on the order of 150 candidate items plus a separate set gathered specifically to detect coverage gaps.

From that pool, around 28 articles are selected and read in full. The rest inform the picture without being cited. Anything that could not be fetched, whether paywalled, blocked, or served through a wire bridge, is used on headline and summary only, and every edition says so explicitly wherever it happens.

Three rules, applied in order.

Original reporting outranks aggregation. A story is credited to whoever did the work, not to whoever repackaged it.

Documentary evidence outranks characterization. A 911 recording, a booking log, a court filing, or a released letter carries more weight than any description of it, including a description offered by the people it concerns.

Argument is labeled as argument. Opinion columns, editorials, and analysis pieces are named as such and never presented as reported fact. When an outlet is arguing rather than reporting, the edition says so before quoting it, and figures cited inside an argument are attributed to that outlet rather than adopted.

Every figure, date, casualty toll, dollar amount, and officeholder is checked against live search at assembly time rather than recalled. Where a claim is load-bearing, it is confirmed against multiple independent outlets or against the primary document, and the edition names which.

This matters more than it sounds. A number that was correct last week is not necessarily correct today, and a name that was correct last year is frequently wrong now.

Government claims about casualties, medical care, enforcement, and fraud are attributed to the government that made them.

One standard, applied regardless of which government is speaking. A health ministry's own case counts are labeled as the ministry's own, whether the ministry is friendly or hostile. A defense department's account of a death in custody is labeled as that department's account. Human rights bodies are credited the same way on a left government as on a right one.

Where two accounts conflict and nothing has been adjudicated, both are reported as conflicting. The edition does not pick a winner it has no standing to pick.

Every edition ends with a two-way comparison. What conservative outlets carried that left and independent outlets did not, and what left and independent outlets carried that conservative outlets did not.

Both directions run at full strength. The self-critical direction is not softened, shortened, or moved to the end. If left and independent outlets spent a day on detention conditions and missed an entire rolling legal campaign, the edition says that, and says which is the harder work and which is the reactive half.

The section also separates gaps from narrative. A story the other side ran that turns out to be sourced to nothing verifiable is not a gap in anyone's coverage. It is intelligence about what an audience is being fed, and it is labeled that way.

A factual claim without a citation is a bug. Vague attributions ("court records show," "reporting has established") without a named source are bugs. In an investigation, every analytical claim carries, inline, the following:

• Primary source identifier. The specific primary source, bracketed inline.
• Page number. Where the specific claim can be verified within the document.
• How the document was found. Search query, entity connection, or thread of discovery, carried in the document record, not the prose.
• Reliability tier with justification. Tracked in the evidence note backing the claim, visible in the article's tier meter.

A bracketed reference appears in the sentence where the claim is made, and links to the record for the source it names. The reader can verify any claim at the point of reading without scrolling or navigating away. An Open Record edition carries the same discipline as numbered footnotes instead, so the outlet and the archived copy travel with the citation.

The Research Vault has been withdrawn. It held document pages built from the 2025 Justice Department release, and too many of them were not ready to publish: uncorrected OCR, records nobody had read closely enough to excerpt, entity pages that were a heading and little else. Publishing the corpus is still the intent. Publishing it in that condition was not.

Citations did not go with it. Every EFTA reference on this site now resolves to the document at its primary source, the Justice Department's own release, and each of those links was checked against the server rather than assumed. A citation is a claim about where something came from, and losing the vault should cost a reader a convenience, not the provenance.

Two things are worth knowing before you click one. The Justice Department puts an age-verification screen in front of the release, so the first thing a citation link produces is a prompt asking whether you are 18, not the document. And the department publishes one PDF per document, named for the Bates number on its first page. A citation to a later page therefore opens the document that contains it, which is why the number in the citation and the number in the filename sometimes differ.

Each citation also carries a second link, marked EE. That is the same document on Epstein Exposed, an independent archive that mirrors the release and publishes its own analysis of individual records. It is there for two reasons. A reader may prefer not to send a justice.gov request across the network they happen to be on, which is a reasonable thing to prefer and not one this publication needs explained to it. And that archive's work on a given document is often worth reading next to the document itself.

The Justice Department link is the citation. The mirror is an alternative host, not a second source: both serve the same scanned pages. One copy was compared page by page against the other and the images match, with the differences confined to re-compression and a separate OCR pass.

Contradictions are treated as their own research object, not as an inconvenience to be resolved by preference. The procedure:

1. Document both pieces of evidence separately, each with its citation and internal tier.
2. Assign Internal Tier 4 (Disputed) to the contradicted claims.
3. File the contradiction in the Contradictions analysis folder.
4. Open the contradiction as its own investigation thread.
5. Do not resolve contradictions by choosing the version that fits the expected narrative.

A contradiction that cannot be resolved is still published. The article names both sides of the disagreement and flags the contradiction as open. The reader is not asked to trust the reporter's preference between competing primary sources.

A large documentary record is the kind of material that will confirm any hypothesis the reporter starts with, if the reporter only looks for confirmations. The discipline against that tendency is procedural, not moral.

• Actively search for evidence that contradicts the working hypothesis. Negative results are part of the record.
• Document searches that returned nothing, with the query and the date, so later work does not repeat the dead end.
• Maintain a running Contradictions file and review it at the start of every session.
• Distinguish between "no evidence found" and "evidence of absence." The first is a gap in the record; the second is a finding.
• When the reporter feels most certain about something, that is the moment to check hardest.

Every EFTA number in the prose is a citation stamp: the identifier in brackets, the page where the claim rests when it rests on one, and a tier glyph. Citations appear in the sentence where the claim is made, not in a footnote block at the bottom, so a reader can check any claim at the point of reading without scrolling or navigating away. Where each stamp points is set out under "Where a citation points" above. Where no verified source URL exists for a number, the stamp stands as text rather than as a link that goes nowhere.

The anchor-documents rail on every article page lists the primary sources the article rests on, with page counts and tiers. That is the article's evidentiary foundation, visible at a glance. The tier bar shows the T1 / T2 / T3 distribution across the piece. A reader who only scans the rail and the bar still knows what the article is built on and how heavily it rests on primary sources.

Every article on this site ends with two sections that are editorial content, not footer metadata: What the Record Does Not Show and Several Items Remain Open. These are the last paragraphs of the piece. They are not concessions and they are not caveats. They are findings.

An investigation that publishes only what it found and hides what it could not find is advocacy. An investigation that publishes what it could not find, with specific document targets and the next retrieval steps, is a case file. This publication ships case files. Gaps in the record are part of the record.

The reader is treated as an adult. The documents are named. The tiers are visible. The contradictions are surfaced. The questions the record cannot answer are published. When the reader follows the citations to the primary sources, the chain of evidence is reproducible.

It does not attempt neutrality between accurate and inaccurate accounts. Marking a government's claim as that government's claim is not the same as treating it as equally likely to be true.

It does not treat volume as coverage. Five opinion columns arguing about a war are not coverage of that war's logistics.

It does not launder argument through the word analysis. Where this publication is arguing, it says so.