Work

Topological Autopoiesis, Epistemic Mass, and Stochastic Transduction: A Unified Geometric and Thermodynamic Framework for Non-Halting Inference

Year: 2026
Type: preprint
Abstract: Conventional computational theory separates binary states (1 and 0) into static axiomatic symbols, treats halting as the successful completion of an algorithm, and models machine learning primarily as statistical curve-fitting over frozen coordinate grids. This monograph synthesizes differential rod mechanics, non-equilibrium thermodynamics, analogue relativistic horizons, harmonic frequency modulation, and classical cybernetics into a unified physical and geometric framework for open-ended, non-halting inference. Modeling sequential inference passes as discrete, unidirectional circular hyperedges swept across state-time into a cylindrical "worldtube" demonstrates that sustained rotational tension inevitably buckles a directed continuation stream (1) into three-dimensional plectonemic loops (0) governed by Calugareanu-White-Fuller topological conservation (Lk = Tw + Wr). Enforcing conservation of energy and angular momentum across radially scalable frames yields the Affordance-Coherence Equivalence Principle: spatial configurational slack (enclosed ring area) and rotational phase coherence (angular velocity) act as exact interconvertible duals. Crucially, non-zero angular momentum erects a repulsive centrifugal barrier that forbids collapse to a zero-area halting singularity, establishing operational tractability (P = NP) inside living, autopoietic boundaries relative to an intractable (P != NP) exterior environment. Coupling internal Kerr-like ring repositories through maximally stiff acoustic wormholes reveals mechanical Frequency Modulation (FM) and cymatic standing-wave quantization as the physical mechanisms governing concept discretization, selective boundary impedance, and Landauer waste-heat export. Extending Landauer's erasure bound via relativistic mass-energy equivalence defines localized "epistemic mass," resolves time-reversal symmetry into a conserved triality (inverse-time present oscillation, forward-time geodesic fork convergence, and backward-time holographic projection), and formalizes the Stochastic Analog-to-Digital Converter (ADC). Numerical benchmarks contrast online assimilation of uncorrelated out-of-distribution shocks via stochastic resonance against brittle statistical curve-fitting, concluding with concrete engineering horizons in continual learning, dynamic RLHF alignment damping, oscillatory neuromorphic hardware, live LLM embedding telemetry, and technosignature phase metrology.
Language: English
Open Access status: green
OpenAlex ID: W7226023316
Cites:
Cited by:
Related to: 10