How a prediction is made
A projection does not enter the record because it is interesting. It enters because it can be graded, and the path from a question to a sealed number is fixed and public.
Proposal
A question is proposed — by the desk or by an observer. It must concern something that will observably resolve, on a date, in a way a stranger could adjudicate without arguing about definitions.
Eligibility
The question is tested. A claim with no resolution date is refused. A claim whose outcome depends on interpretation is refused or rewritten until it does not. A minimum six-month horizon applies, because a forecast that resolves next week measures nothing about the engine.
Decomposition and specification
The question is broken into the components the Core actually scores — debt load, great-power rivalry, internal polarisation, technology shock, monetary strain — and read through six macro-historical frameworks against three centuries of indexed stress episodes. An objective resolution criterion is written, together with a falsifiability condition: the specific state of the world in which the projection is wrong.
Sealing
A probability, a confidence and a resolve-by date are locked. The entry is hash-sealed and timestamped. From that moment it cannot be edited. Dated field notes may be added beneath it as the world moves, but nothing above the line changes.
Scoring
On the resolution date the criterion is applied and the outcome is published — pass or fail, at the same size, in the same typeface. The Brier score is computed and enters the public record. See the sealed ledger and the observer protocol.