Recursive prediction under incomplete evidence.
Learned Evolution explores predictive engines that evolve state through time when forecasts may recurse between incomplete or asynchronous observations and later evidence can update the system's best current estimate.
A native engine, not the Instrument itself.
Learned Evolution is developed alongside the ABSLTN Predictive Instrument, but the Instrument is model-neutral. Existing predictive systems can be assessed without using Learned Evolution.
Study behavior when the system cannot continually reset from complete ground-truth observations.
Later evidence can update the current estimate without rewriting what was previously predicted.
Current public evidence concerns architecture and controlled synthetic pressures, not universal predictive superiority or live-domain production performance.