How a record is made

Every entry in the ledger is a receipt for one claim: the claim in canonical form, a verdict, two grades, and the evidence they rest on. This page explains where each part comes from, what can never influence it, and how a record changes over time.

The verdict: which way the evidence points

A verdict is a statement of direction, nothing more. There are exactly four:

• Supported. The weight of evidence points toward the claim.Decisive support share of 55% or above.
• Contested. Substantial evidence points both ways.Decisive support share from 46% to 54%.
• Unsupported. The weight of evidence points against the claim.Decisive support share of 45% or below.
• Insufficient record. There is not yet enough qualifying evidence to state a direction. This determination comes first: if the evidence base does not meet our sufficiency requirements, no direction is stated at all.

Each term is bound to published numeric thresholds over the support and contradiction structure of the evidence, shown above. The thresholds are calibrated against a fixed suite of claims with well-established status, spanning settled findings, collapsed findings, and live controversies, so that a settled claim cannot render as doubtful and a collapsed one cannot render as supported.

The two grades: how strong the ground is

A verdict tells you which way the evidence points. The grades tell you how solid the ground is under it. There are two, and they answer different questions.

• Evidence grade (A to F) Measures the evidence against a single strict standard applied identically to every claim in every field: replication, independence of the groups involved, methodological quality, and the stability of results. It never adjusts for any field's customs. A weak evidence base grades low no matter how popular its conclusion.
• Field-relative grade (A to F) Measures the same evidence against the strongest form evidence can realistically take for that kind of question. Some questions permit controlled trials; some can only ever be studied observationally; fields differ in how much replication their structures produce. The ceiling for each field is derived empirically from the literature itself, not chosen by us, and every field-relative grade carries a basis line stating what it was measured against.

Reading them together is the point. A claim can be the best its field can offer and still rest on limited ground; the two grades say both things at once instead of hiding one inside the other. Grades are letters, never percentages, because a percentage reads as the probability a claim is true, and that is not what a grade measures.

The evidence

Every published record lists its sources: title, authors, year, venue, and a resolvable identifier, in public, on the record. Every citation on every record is machine-verified to resolve to a real publication with matching title and year before the record can publish. A record whose sources cannot be verified does not ship.

What can never touch a score

Money moves priority, never outcome. Whoever pays for a verification, and whether anyone pays at all, has no path into the verdict or the grades, and every record states who commissioned or sponsored it.

Public opinion carries zero weight. We measure what people believe about claims, but that data is stored where the scoring system has no ability to read it, and we prove it continuously: the pipeline is run with that data present and with it removed, and the verdict and grades must be identical to the last bit, or the system fails its own tests.

The classification of a claim into its field is printed on the record, so the standard it was graded against can be inspected, not taken on trust.

How records change

The ledger is append-only. A published record can be added to, never rewritten and never withdrawn. When new evidence changes a verdict or a grade, the change is made in public, dated, with the triggering evidence named, and the record's full history remains visible. We treat corrections as a feature of the record, not an embarrassment to it: a ledger that cannot change its entries cannot follow the evidence.

How claims enter verification

Entries advance through continuous prioritization across three paths: editorial selection, reader demand, and sponsorship. Commissioned work is private by default and publishes at the customer's choice, with two permanent rules. A displayed verification mark always requires a published record. And privacy attaches to the engagement, never to the claim: no one can prevent a claim itself from being verified through public demand, and once a record is published, it is published forever.

What we do not do

We do not decide what is true. A record states which way the evidence points and how strong the ground is; the conclusion remains yours. Responsibility for decisions stays human. We provide the record that makes them defensible.