The instrument locks a judgment, watches the world, and keeps the score.

A claim moves through five states, and only forwards.

A claim is one sentence from a published report about a resolvable outcome, with the probability the firm puts on it. It starts as a proposed candidate, extracted from the report and never shown. It becomes locked when it passes the three tests below: the statement, the probability, the date it was made and the window it resolves in are then fixed and never edited. When its window passes without a public source it becomes open, and an open claim adds nothing to the score. A dated public source resolves it true or false. A claim that proves unresolvable as worded is void: excluded from the score and counted separately, so the reader can see how often the firm wrote a claim that could not be settled. Where sources of equal standing disagree, the claim is contested: held open with the contest written into its note, until the founders rule.

What qualifies

A statement in a brief becomes a locked claim only when it passes all three tests. First, the brief states a probability for the outcome, or attaches a confidence band to the judgment that carries it. Second, the outcome is checkable from public sources: a filing, a published series, a dated official record, or reporting that meets the source tier policy in Chapter 02. Third, the resolution window is stated in the brief or is unambiguous from it. A statement that fails any test is kept on file as a proposed claim, is never shown publicly, and is never scored.

Two cores: one remembers, one reasons.

The Mind is the memory the instrument reasons from, kept as four kinds of knowledge that are never blended. Semantic memory holds what is true in general: base rates for defined reference classes, each with its sample size, its window, its sources and a date by which it must be re-grounded; and principles about how actors in a domain behave, each traced to its source work and held in tension with the principles that contradict it. Episodic memory holds what the firm observed and scored on its own work, dated and, where it is client-attributed, firewalled from everything shareable. Procedural memory is the method itself, AEGEAN-OS, which the machinery may read and never edit. Working memory holds the one live problem. Scored outcomes sharpen the priors over time; a named human approves every such change, and none of it may touch the doctrine.

Metron is the reasoning core. It reasons from the Mind rather than from the model alone: it anchors a case on the reference class it belongs to, weighs the base rate against what the firm has itself observed and scored, and runs the futures, including an opponent pursuing its own goals. What comes out is a calibrated judgment with a credible interval whose width is honest, every number traced to a source and every judgment to a principle, so the reasoning can be cross-examined line by line. How the weighing and the adjudication are done is the firm's method and is not published.

Both cores are suggest-only. The runtime cannot reach an approve surface; a named human locks the claim and resolves it. That is the first standing rule, and every other rule follows from it.

The machine-readable fingerprints, record classes, selection counts and content-addressed ledger snapshots are published in the Record integrity area.