Sealed forecasts

Models on a sealed forecast

When a run records them, the seal page says which model was frozen for each role at the start of that run. The line is that run's record. It is not a standing claim about which model runs next.

Simulated by the simulation model; planned and read by the planning model; lines written by the dialog model when that was someone else; outcomes judged by the synthesis model.

• Planning

  Classifies the question and plans the run. This is the "planned and read by" name.

• Simulation

  Plays the events and the blind replay. This is the "Simulated by" name.

• Dialog

  Writes the staged lines. Those lines are simulated, not quotes. Named on the seal when it differs from the simulation model.

• Synthesis

  Writes the final analysis. This is the "outcomes judged by" name.

The public seal does not yet carry the date a default changed, or the taxonomy version for that run. Those stay off the line until the seal records them.

Simulations are labeled as simulations. We don't publish accuracy claims.