Who publishes this,
and how it is checked.

What this site is

Do Your Own Research (“DYOR”) publishes long-form research dossiers on claims that are widely held, widely disputed, and badly served by both the summary and the counter-summary: the September 11 attacks, the Kennedy assassination, the moon landings, Project MKUltra, the Bavarian Illuminati, flat-earth belief, and sterilization allegations against WHO-linked vaccination campaigns.

Each dossier is the record of a single research run. It carries a bounded question, the competing explanations that question was tested against, the verbatim passages that were actually read, how each passage bore on each explanation, the attacks the run made on its own leading conclusion, the confidence its own ledger computed, and everything it could not settle. The complete unedited run transcript — typically one to three thousand lines — is published alongside every dossier as a downloadable file.

The site is a static publication. It has no accounts, no logins, no newsletter, no comment system and no forms. Every dossier remains fully readable with JavaScript restricted, cookies disabled and an ad blocker running.

Why it exists

The internet is not short of answers about any of these topics. It is short of answers you can take apart.

The failure mode this publication is built against is the easiest one in knowledge work: collecting only the information that supports the answer you already like, then presenting the result as a verdict. Opening with “unsupported, confidence high” repeats exactly the structure that made a claim contested in the first place — an institution asking a reader to accept its conclusion. That works when the reader already grants the institution. It fails precisely where it would be most useful, because a claim wrapped in distrust cannot be dissolved by a summary that sounds more certain.

So the useful output is not a better pronouncement. It is a path the reader can walk: origin, retelling, prediction, test, observation, contradiction, and only then a conclusion — if the evidence supports one. A reader who disagrees should be able to point at the exact step where their judgment diverges from the record. Disagreement, on this site, is supposed to have an address.

How a dossier is made

Every run passes through the same five stages, in the same order, before anything is written for publication.

• 01 — Bound the question. Define the population, period, geography, terms, and what is deliberately out of scope, before an answer can quietly move the boundary. The scope is published at the top of the dossier, and it says plainly what the run does not address. The flat-earth run, for example, placed the shape of the Earth out of scope before searching; nothing in that record bears on it either way, and the page says so.
• 02 — Compete explanations. The leading story is given a real opponent. Each explanation is written out in full, and each is paired with the observation that would disprove it. The run also records the explanations that no available count could settle.
• 03 — Freeze the evidence. Every source is snapshotted and content-hashed. Every quoted passage must survive a verbatim gate against the stored bytes before it can be cited. The SHA-256 digest of the snapshot is printed beside the source in the published dossier, so a quotation can be checked against the bytes that were actually read rather than against whatever the URL serves today.
• 04 — Attack the leader. A separate, separately budgeted falsification pass runs adversarial counter-searches against whichever explanation is currently winning, and stress-tests each asserted finding with a named failure condition — the specific thing that, if found, would break it. The searches that came back empty are published as empty. An empty adversarial search is not evidence of absence, and the dossiers say so in those words.
• 05 — Show the uncertainty. Confidence is computed from the evidence ledger rather than typed into a paragraph, and the limitations, contradictions and open questions are printed in the main reading path — not tucked into an appendix.

What is inside every dossier

• A bounded question. Population, period, geography, definitions, assumptions, exclusions, and the ambiguities the question itself recorded.
• An evidence ledger. Verified verbatim passages tied to content-hashed snapshots, then scored against every live explanation in a bearing matrix. Passages that nothing scored are listed too.
• The case against it. Pre-registered probes, contradictory evidence, alternate explanations, and the recorded attacks on the run's own leading conclusion.
• What remains unknown. Missing evidence, untraced provenance, subquestions no passage answered, audit measurements printed even when the result is zero, and the next questions.

The intended shape is a piece that is readable in minutes and auditable for hours. The story orients you; the evidence record lets you check every consequential step.

What is machine-made, and what is not

This should not be buried, because it changes how the dossiers ought to be read.

The research runs are produced by an automated pipeline. It performs the searches, snapshots and hashes the sources, gates every excerpt against the stored bytes, scores each passage against each explanation, runs the falsification pass, and computes the confidence level and the evidence score. No stage is asked to label its own confidence, and no model types one: the levels are derived from counts in the ledger.

The published pages are then written from that record for a reader. What the editing does is select, order and explain. What it does not do is change the run's numbers, its confidence levels, its findings, or the direction in which its evidence pointed. Where a dossier's page states a count — “ten passages bear strongly against”, “zero of twelve snapshots reached an upstream origin”, “14 out of 100” — that number is the run's own, and the full transcript published with the dossier is the check on it.

Where a run recorded no human reviewer approval of a stage, the dossier says that in its limitations. Nothing here should be read as a claim that a person independently verified each step.

What this site is not

• Not a news site. There is no feed, no topical commentary and no publishing schedule to fill. The archive is deliberately small — fewer reconstructable dossiers rather than a stream of summaries.
• Not an AI answer engine. The output is a method transcript with the evidence attached, not a generated response to a prompt.
• Not advocacy. DYOR is not affiliated with, funded by, or acting for any government, party, campaign, advocacy group, advertiser or the subject of any dossier.
• Not a verdict machine. Several dossiers deliberately reach no run-level verdict. That is a result, not an omission, and the editorial standards explain precisely what it does and does not mean.
• Not medical, legal or financial advice. Where a dossier touches a medical claim it is a record of what specific documents state about that claim's evidentiary basis. It is not guidance about your health, and it is not a recommendation to accept or refuse any treatment.

Who is accountable

Do Your Own Research is an independently operated publication, published under its own name rather than a personal byline. The publisher is responsible for every page on this domain: the selection of claims, the editing of each dossier, and any error in either.

Accountability here is meant to be structural rather than reputational. A byline asks you to trust a person. The three things below let you check the work instead:

• The transcript. Every dossier links its complete, unedited run record, published as a file on this domain.
• The digests. Every quoted passage in an empirical dossier is printed beside the SHA-256 digest of the snapshot it was read from.
• The commit history. This site is a public repository. Every edit to every page, including this one and the corrections log, is a public, dated, attributable change.

If you believe something on this site is wrong, there is a stated route for saying so and a stated commitment about what happens next. It is on the contact page, and the policy behind it is in the editorial standards.

How it is paid for

DYOR takes no sponsorship, no paid placement, no affiliate commission and no payment of any kind from any party with an interest in what a dossier concludes. It publishes no guest posts and sells no links.

Its intended revenue is display advertising served by Google AdSense, disclosed in full on the privacy and advertising page. Advertisers have no involvement in, and no advance sight of, what is published; advertisements are labelled and are placed outside the body of a dossier — never inside the argument, the evidence ledger or the source list. No advertisement should be read as an endorsement by us or of us.

Getting in touch

General correspondence, corrections and disputes: contact@doyourownresearch.me. Privacy and advertising questions: privacy@doyourownresearch.me. The contact page sets out what to include when you are challenging a specific finding, and what we do with it.