State-Dependent Informational Substitution in Memory and Prediction

This preprint introduces the SIDSMP framework (State-Dependent Informational Substitution in Memory and Prediction), a constraint-based dynamical model in which predictive structure emerges, stabilizes, and collapses under finite informational and energetic budgets. The model is substrate-independent and does not assume specific neural, biological, or artificial implementations. Instead, it formalizes minimal structural constraints governing transformability, structural work, dissipative cost, and predictive capacity in finite informational systems. The present work is foundational and theoretical. It provides a formal structure, identifies admissible dynamical regimes, and defines falsifiable predictions, while leaving domain-specific operationalization and empirical calibration to future research. A reproducible computational toy model implementing the validation suite is archived separately and linked under Related Identifiers.