Sources and methodology
The verification protocol
Every factual claim on this site is born flagged. The flag reads TODO-VERIFY and it renders in public. The flag comes off when, and only when, a source has been fetched, read, and found to carry the claim, and the words that carry it are recorded in the data file next to the link. That is the whole standard.
- Sourcing. Receipts are assembled for each flagged claim: exact quotes, dates, statute or bill numbers, page citations, and links. Paraphrase without a receipt does not count. Claims about events after January 2026 must come from post-cutoff sourced research, never from model memory.
- The source check. The source is retrieved and read, not merely linked, and the passage that carries the claim is quoted into the record along with the date it was read. If nothing in the source says the thing, the flag stays up and the gap is published.
This changed on 2026-08-12, and the change is worth stating plainly
Until that date this page described a two-stage pipeline whose second stage was a named human verifier signing
off. That gate is gone, at the site owner's decision, on the reasoning that nobody holds a crystal ball for
ultimate truths and a source that says the thing is the standard a reader can actually check. So the honest
description of what a [sourced] chip on this site now means is narrow: a source was read
and it says this. It does not mean a human adjudicated the claim, and it does not mean the source is
right.
What replaced the human gate is a harder artifact rather than nothing. The old rule asked who signed; the new rule asks what the source says, and the build refuses any claim marked verified that does not record the quoted words, a source, and the date it was read. A URL on its own is explicitly not enough, and that is not a hypothetical: four sources on these sites resolved perfectly well while carrying nothing. Two US State Department pages returned HTTP 200 and served a page titled "Technical Difficulties." A Xinhua page correctly titled "Full text of Constitution of Communist Party of China" contained no constitution, only a pointer. An archive snapshot of a newspaper article captured the paywall instead of the article. And one claim on this site's own rhetoric page cited two sources for a quotation that neither of them contained, which is why that claim now carries an UNRESOLVED note instead of the quotation.
Two tiers, because not every claim is the same kind of claim
Checking that a Library of Congress country study prints a particular date is not the same job as weighing scholarly estimates of excess mortality. Since 2026-08-11 every claim carries a tier that says which job it is. Tier A claims assert only dates, names, discrete events, or quoted wording, so a reader holding the source could confirm them by reading, without weighing anything. Tier B claims require judgment, and additionally record what that judgment was. The tier governs what a claim's receipt has to show, never how well it has to be evidenced: both tiers need a quote, a source, and a date the source was read.
The current split is 18 claims in tier A and 105 in tier B. That ratio is worth stating plainly, because it is the honest answer to why so much of this site is still flagged: roughly six sevenths of what this site asserts involves a judgment, and judgments cannot be checked in bulk. The build enforces the distinction rather than trusting it. A tier A claim whose wording contains judgment language fails the build, and a claim marked verified with no quoted passage fails the build too.
The site has five data files, and they do not all record verification the same way, so here is exactly what each
one carries. countries.json (32 entries) and variants.json (15 entries) carry a full
verification block on every entry: status, sources, verifier, and date. record.json (13 entries) and
rhetoric.json (30 entries) carry a status and sources on every entry, but no verifier or date fields
yet. misconceptions.json (10 entries) carries sourcing and a status on every individual receipt
rather than on the entry as a whole. Two entries carry an empty sources list: the democratic socialism and social
democracy comparators on the variants page, which have no key dates of their own to verify because their
definitions are owned by the sister sites. Separately, and true of all fifteen variant cards equally, the
seven-axis scores are this site's own placements rather than sourced facts, and as of 2026-08-12 they are
labeled that way rather than flagged as pending. There is no source that could carry "coercion: 3", so leaving
them under a TODO-VERIFY flag implied a check that was never going to arrive. The anchors for each score are
printed on the spectrum page so the reasoning can be checked instead of a citation. Bringing the last three data files onto the full verification
block, and putting the axis scores on a defensible footing, is outstanding work.
Editorial rules are enforced structurally where possible: the build fails if Nazi Germany is entered as a
communist state, because that classification is settled (see the
misconceptions page, M-01). The structural guards are strongest on
countries.json, which is checked against the full schema and the editorial rulings; the other four
files are checked for sourcing, status, and tier vocabulary only.
One of those guards was broken and is worth recording rather than quietly fixing. The rule that nothing reaches
"verified" without its receipt recorded was, until 2026-08-11, enforced everywhere except
countries.json, the most heavily checked file on the site. No entry had been wrongly cleared, because
no country entry has been cleared at all, but the guard would not have caught it. It was found by deliberately
feeding the checker broken data rather than by reading the code, which is now the practice whenever a guard
changes.
Method commitments
- Ranges, not cherry-picked numbers. Where scholars dispute a figure, especially death tolls, this site presents the range and the methodological debate, not a single convenient number.
- Debunking runs both directions. The same standards apply to "everything I dislike is communism" and to "real communism has never been tried."
- Contested classifications carry the dispute. North Korea and post-reform China are presented with the scholarly disagreement visible, not resolved by fiat.
- Shared definitions. The definitions ladder is identical across the four sister sites and changes only in lockstep.
Bibliography
The bibliography accumulates as claims are sourced. Sourcing began on 2026-08-11 with three claims on the home
page, under the older protocol that required a human sign-off; those three were checked by the site owner
against Marx's Critique of the Gotha Programme, Kolakowski in Socialist Register, and the rhetoric corpus, and
they stand. From 2026-08-12 the standard is the one described above, and a cleared claim renders a
[sourced] chip carrying the quoted passage and linking to the source it came from. Everything not
yet cleared still carries its [TODO-VERIFY] flag, and at the time of writing that is still most of
the site.