Canonical model run
Market-data date, computation timestamp, model version, state, allocation and source digest are stored together.
Every observation begins with a real canonical run, carries its model version and timestamps, and remains private for a separate 14-calendar-day embargo. Hypothetical backtests never enter this journal.
A simulated historical row can never be relabelled as a forward observation. New rows are append-only and become visible only after their own publication timestamp.
Market-data date, computation timestamp, model version, state, allocation and source digest are stored together.
The observation exists, but the public API and CSV withhold it until publishAt is reached.
A released row stays clearly identified as delayed and is never blended with hypothetical proof.
Only rows whose public-release timestamp has passed can appear below.
Counsel-reviewed operator and producer-disclosure approval is not complete, so public signal states remain withheld.
The proof lab tests historical assumptions. This journal records what the versioned model produced after tracking began. Each has a different evidentiary job.