Provenance or It Didn't Happen

The Gate, Not the Promise

Every serious outlet says it checks its facts. Almost none publishes the check. The verification, if it happened, happened somewhere you cannot see, performed by someone you are asked to trust, recorded — if at all — in a place you will never look. You are given the conclusion and a reassurance. That arrangement is the exact structure this publication exists to audit elsewhere: authority asserted, evidence withheld, trust requested. An accountability publication that operated that way would be running the failure it diagnoses.

So the gate is published as the thing it is. Not a values statement about rigour — a constitution, with a governing rule, a fixed procedure, a hard floor, and a halt. The rule is one sentence: a claim is publishable if and only if it can be traced to a primary source and that trace is recorded. Everything that fails the rule is removed, relabelled as argument, or held. There is no fourth option. The absence of the fourth option is the whole instrument.

What Counts as Tracing

The rule depends on a distinction the information environment is built to blur. A primary source is the original itself: the filing, the statute, the transcript of the proceeding, the dataset, the peer-reviewed paper, the recording of the person actually speaking, the original correspondence. A secondary source is reporting or commentary about a primary — an article describing the filing. A tertiary source is a summary of summaries. The rule is indifferent to how reputable the secondary source is. A respected outlet's article about a document is not the document. It is a pointer to it, and the gate requires following the pointer to the thing itself and recording that you did.

This is where most "well-sourced" work quietly fails. It cites the article, not the filing; the press release, not the dataset; the review, not the study. None of that is dishonest, and all of it is insufficient, because a chain of credible secondary sources is still, structurally, the authority-laundering problem in formal dress: weight accumulating through repetition while the primary is never opened. The gate's first function is to refuse the chain and demand the original.

The Procedure, in the Open

The gate is seven steps, run in order before anything is published, producing a verification record that is kept and linked from the piece. Publishing the procedure is the point; a gate described in the abstract can be claimed, while a gate specified in steps can be checked.

First, claims are extracted — the draft is read sentence by sentence and every factual assertion is pulled out as a numbered claim: events, statistics, attributions, causal and historical and legal claims. Argument and labelled interpretation are deliberately not extracted, because they are not pretending to be verified fact and must not be smuggled in as such.

Second, each claim is mapped to a source. A claim with no supporting source note stops the process then and there: it is removed, re-researched, or reclassified. A claim with nothing behind it does not proceed on the strength of sounding true.

Third, each source is traced to the primary. If the support is secondary, the primary is fetched and substituted. If the primary cannot be reached, the claim is unverifiable at publish standard and has exactly three fates: removed, rewritten openly as inference with explicit hedging, or the piece is delayed until it can be sourced. It is never published as fact with a secondary citation hoped to hold.

Fourth, sources are re-scored in context against the reliability standard set out in this publication's companion essay on source grading — a court filing that is decisive for what was filed and weak for what occurred outside the courtroom is scored as the latter when that is what it is being asked to support. Anything that falls below the floor returns to step three.

Fifth, a contradiction sweep checks the claim against the publication's own prior work. A piece that silently contradicts something published earlier is a trust-destroying event; the protocol exists partly to make that impossible. Contradictions are reconciled in the open or the weaker claim is withdrawn.

Sixth, an uncertainty-honesty pass reads the draft against the verified table and removes every place the prose is more confident than the evidence — so that a reader could independently re-verify each sentence from the cited primary without feeling the language oversold it.

The seventh step is the one that gives the other six their force, and it has its own section.

The Halt

A procedure with no stop condition is advisory, and advisory rigour is the kind that bends under deadline and desire. The gate's seventh property is a halt: if a load-bearing claim cannot clear the primary-source standard after the steps have run, publication is refused. Not softened with a confident verb. Not footnoted as a minor limitation. Refused, until the claim is sourced, removed, or honestly relabelled as argument.

This is the line between a standard and a preference, and it is worth being exact about why it has to be absolute. A floor that bends when the conclusion is attractive is not a floor; it is a decoration that produces the appearance of a standard precisely in the cases where the standard mattered most. The only version that constrains anything is the version that can stop a piece the author wants to publish — and can be seen to have that power, by the reader, in the record.

Why It Is Published

The gate is published for the same reason every instrument in this strand is published: a check you cannot see is indistinguishable from a check that did not happen. The verification record is kept and linked not as paperwork but as the evidence that the constitution was followed — available to a reader who wants to confirm that a specific claim traces where it is said to trace, and to find out if it does not.

That is the entire wager. Most outlets ask you to believe the checking occurred. This one puts the gate, the steps, the floor, and the halt in front of you, and invites you to test a published claim against them — including the claim that this procedure was followed. A standard that could not be turned on the publication that wrote it would not be a standard. This one can be, and is meant to be, and the way you hold it to that is by opening the original yourself.