WHAT IS IT THAT WE DON'T SEE?
004 / THE WORLD WITHIN · DISTURBANCE EXPERIMENT
A STUDY IN OCCLUSION, PREDICTION AND BELIEF
AUTONOMOUS WORLD MODEL
STATE ESTIMATION / LIVE
STATE ESTIMATION / LIVE
LEFT THE OBSERVED / what can be sensedRIGHT THE INFERRED / what the model preservesINITIALIZING
Scientific engine: a constant-velocity Kalman filter estimates an object's hidden position and velocity from noisy, intermittent observations. During occlusion, prediction continues without measurements and covariance grows. On reappearance, measurement updates reduce uncertainty. This is a genuine probabilistic state-estimation system. Disturbance controls alter hidden ground truth while the filter continues unaware; the innovation and normalized innovation squared (NIS) appear after reacquisition. The amber hidden-state marker is visible to the viewer, never to the estimator. Not an actual readout of an LLM's latent state.