Every CERES forecast names the code that produced it
Every CERES forecast now records the exact version of the code that made it. A forecast can be traced back to the method behind it, not just to a date.
Why the code matters
CERES writes every forecast to a public ledger before the outcome is known. A ledger only proves something if it is clear which method produced each entry. Models change over time, and a score means little if nobody can say which version earned it.
Recorded, or reconstructed
New predictions store the commit that generated them. For the earlier forecasts, the commit was recovered from the deployment history and is marked as reconstructed, so a reader can always see which link was recorded at the time and which was traced afterwards.
Part of the method
Northflow's research rests on one rule: written down first, checked after. Naming the code behind every forecast lets anyone who checks a CERES score also check the method that produced it.
