Adaptive Multiscale Information Encoding and Decoding on Fractal Manifolds

We present a self-contained framework for encoding and decoding discrete information on a multiscale manifold induced by Julia dynamics, designed to operate under unknown channel distortion. Information is embedded as structured phase perturbations across concentric potential shells, with decoding performed via cell-level aggregation and hypothesis testing on dual, incommensurate reference fields. A central challenge addressed in this work is interference cancellation when the reliability of one reference field is itself uncertain. To resolve this, we introduce a deterministic, validation-budgeted adaptive cancellation mechanism that selects shell-dependent cancellation weights from a finite candidate set using margin-weighted error estimates and a blur-aware prior derived directly from the received signal. The procedure requires no training data, no external calibration, and no assumed knowledge of channel parameters, and is fully reproducible under fixed seeds. We document the iterative design process leading to the final system, including explored and rejected alternatives such as fixed cancellation, learned predictors, and pilot-biased selection, to clarify the tradeoffs inherent in adaptive decoding under partial observability. The system is not proposed as a practical communication scheme; it serves as a concrete, reproducible example of multiscale signal recovery using only structural priors and internal self-assessment, and as a testbed for future work on self-calibrating systems operating under unknown distortion. No claims are made regarding optimality, capacity, or performance relative to established coding schemes.