Methodology.

Each entry begins with a dossier assembled from the public record: court opinions and filings, peer-reviewed research, inspector-general reports, statutes, campaign-finance databases, official agency publications, FOIA-derived datasets, and named-byline investigative journalism at tier-1 outlets — the New York Times, Wall Street Journal, Washington Post, Reuters, Associated Press, Bloomberg, ProPublica, NPR, PBS Frontline, The New Yorker, CBS News, CNN, and their equivalents.

No single aggregator is ever cited in isolation. Wikipedia, social media, and unsigned blogs never appear on a source strip — only on the cutting-room floor of the research process, as pointers to the primary sources that do make it through.

Every entry has a canonical data register where each scalar, each record, and each array is annotated with two things: a confidence grade (one of “verified,” “reported,” or “inferred”) and a reference back to the section of the research dossier it came from, alongside the outlets or filings the dossier itself cites. No panel derives a number at render time. No headline ratio is composed inside a page component. The data register carries the provenance, and the page renders the data register.

An automated check reads each register before every release and enforces the shape: every record must carry a confidence grade, every dossier reference must resolve, every external source must map to a named, dated outlet or filing. When provenance drifts, the release is blocked.

Casualty counts, strike-damage assessments, shooting counts, and mortality figures are never carried on a single source. They require a second, independent source of comparable tier. When the two sources disagree, both are shown. The public record of the Iran strike-damage assessments, for example, pairs the leaked DIA assessment against the Pentagon's on-record statement; the ICE shooting tally pairs the Trace/Business Insider/Type Investigations FOIA baseline against the WSJ/NYT/AP/Reuters running tally. When one party to a dispute is a government agency and another is a peer- reviewed study, both are cited — never collapsed.

Named individuals appear in entries under two conditions. First: public figures — elected officials, senior appointees, corporate executives — acting in their official capacity, whose conduct is documented in filings, sworn testimony, official press statements, or named-byline reporting at tier-1 outlets. Their names appear because their conduct is the public record.

Second: on-record critics whose statements are themselves attributed — former directors, congressional witnesses, court-cited experts. Their names appear because their words are already in the public record, and their characterizations are rendered as quotes attributed to the speaker, not as independent assertions of this publication.

Individual low-rank actors — a specific guard in a custody death, a specific officer in a shooting — are not named unless the attribution is both tier-1 and multi-source. A single-outlet allegation or an "allegedly per court filings" characterization is not sufficient. When the public record names the position but not the individual defensibly, the position is what appears.

The governing U.S. standard for public figures is New York Times Co. v. Sullivan, 376 U.S. 254 (1964): recovery requires proof that a statement was made with actual malice — knowledge of falsity or reckless disregard for the truth. The methodology above is organized around making that standard structurally hard to trip. Every assessment-level claim is quoted from a named speaker or attributed to a named document. The publication does not declare, in its own voice, that a named individual committed a crime, made a false statement, or acted with a particular intent. It shows what the courts, the regulators, the peer-reviewed studies, and the named-byline reporting have documented, and it lets the reader draw the conclusion.

This is not a legal opinion and is not legal advice. It is an editorial discipline. Anyone considering legal action against a publication would be well advised to consult counsel; anyone considering publishing under a similar posture would be well advised to do the same.

This publication has no author, no byline, no contact form, no mailing address, no analytics pixel, and no advertising. The masthead reads verum. press. because the work is what the work is. Raw research materials are never checked into the public surface of the project. The deployed site carries no identifying strings, no tracking infrastructure, and no referral headers; a Content-Security-Policy locks every asset to first-party origin and an X-Robots-Tag asks the archive bots to skip caching. These are deliberate choices, and they are audited before every release.

If you believe something on this site is wrong — factually wrong, not merely uncomfortable — the correct path is to bring a source of at least equal tier to the one carrying the disputed claim. The publication has no audience-management workflow, so there is no inbox; but the public record is the public record, and the data layer is readable. When a tier-1 outlet reports that we carried something in error, we correct the data layer, mark the record, and ship the correction in the next release.