One-Step Permanent Stability
Abstract
A prediction partition stable for one step remains stable at every later depth.
Theorem 1.1 (One stable step makes every later prediction relation equal).
Proof. Machine-checked in Lean as D5/S3/ObserverMemory/PredictionCertificates/OneStepPermanentStability.one_step_stability_is_permanent (✓ std3). ∎
Source. Repository-derived.
Commentary.
For a self-map F and readout q, ReadoutWord(F,q,m,y) records the readouts of y at update times zero through m. The premise states that equality of words at depths m and m+1 defines the same relation on states.
The first public conjunct says that the depth-m relation is preserved by updating both states. The second says, for every natural offset r, that equality at depth m+r is equivalent to equality at depth m.
Repository search found prediction_partition_stable_forever with exactly this premise and both conclusions. The Lean wrapper imports and applies that declaration directly, without introducing another prediction-word or relation primitive.
The imported module also compiles a constant Boolean readout witness for the premise, so the hypothesis is satisfiable on an inhabited, nontrivial state carrier.
References
- Truth anchor:
D5/S3/ObserverMemory/PredictionCertificates/OneStepPermanentStability.one_step_stability_is_permanent - Dependency: D5/S3/ObserverMemory/Prediction/PredictionPartitionStability