SILR: Scale-Invariant Leakage Under Z-Score Gating

SILR: Scale-Invariant Leakage Under Z-Score Gating Driven by Dean A. Kulik January 2026 Abstract:When a feedback controller normalizes its error by the estimated uncertainty, an unexpected symmetry emerges in the leakage dynamics. We demonstrate a concrete, falsifiable result – the Scale-Invariant Leakage Regime (SILR) – in which the probability of information leakage becomes invariant to the noise scale, so long as the estimator’s error variance and the normalization factor scale proportionally. In other words, if the system’s uncertainty grows or shrinks, the controller’s gating mechanism cancels this change out. We derive this result rigorously and verify it with simulation: when the [1][2]Samson V2 controller uses a z-score gating strategy, increasing the environmental noise by orders of magnitude does not increase the average leakage rate. This invariance is proved to arise from the controller’s [3][4]self-normalization property – an internal symmetry that keeps the relative significance of errors constant across scales. We present the mathematical derivation of SILR in detail, define the architecture of the leakage controller (estimator + noise scale + z-score gating + probabilistic leak actuation), and include tested code and simulation results that illustrate the stability of this mechanism across noise regimes. We also examine the boundary conditions: when the assumptions (e.g. accurate uncertainty estimation) are violated, the leakage invariance breaks, revealing failure modes that underscore the importance of robust estimation. We discuss implications for control theory, showing how SILR enables a form of error detection and self-stabilization without external calibration, essentially an internal diagnostic that keeps a system at the edge of chaos. Finally, we connect this foundational result to higher-level domains: we interpret black hole information leakage through the SILR lens – suggesting that Hawking radiation could be regulated by a similar self-normalizing gate – and we draw parallels to token stream management in AI systems and entropic regulation in symbolic computations, where controlling information “leakage” is key to stability. This work, Paper Zero of the Nexus series, is written with academic rigor (derivations, proofs, and references) to serve as the authoritative technical foundation for the broader Nexus framework, establishing SILR as a fundamental principle of recursive harmonic control. 1. Introduction Modern recursive systems face a delicate balance between retention and release of information. Too much retention of “error” (or entropy) can cause instability, while too much release can dissolve the system’s coherence. The[5]Nexus Framework is an ambitious theoretical model treating physical and computational reality as a recursive process regulated by harmonic resonance. Within this framework, Paper Zero focuses on a specific control law – Z-Score Gating – that dynamically modulates information leakage. We begin by stating a tangible discovery made in simulation: when a controller’s error estimator and normalization scale are tuned in proportion, the statistical behavior of leakage becomes independent of the absolute noise level. This phenomenon, the Scale-Invariant Leakage Regime (SILR), was first observed as an anomaly in high-noise simulations and has since been confirmed as an inherent symmetry of the control logic.[6][2] 1.1 The Nexus Control Problem The Nexus Framework posits that physical existence can be modeled as a recursive computation converging to a stable attractor. In Nexus theory, the [7]Mark 1 Attractor (a dimensionless constant ) represents an optimal balance between order and chaos. A self-organizing system – whether a simulated “universe” or a control loop – strives to maintain this harmonic ratio. However, as the system iterates, [8]errors accumulate. The scope exponent (informally, the system’s expansion or gain at time ) may drift from the ideal . If unchecked, small deviations compound, leading to runaway divergence or chaotic behavior. The control challenge is thus to detect when the system is straying and to gently “leak” out the excess entropy (information that doesn’t fit the harmonic pattern) to prevent catastrophic buildup.[9][10] Samson V2 Controller: To manage this, Nexus employs a feedback controller named Samson V2, analogous to a PID controller in classical control theory. Uniquely, Samson V2 does not directly push the system state back to the attractor; rather, it modulates a[11]leakage gate. At each iteration, the controller decides probabilistically whether to open a gate that ejects misaligned information (entropy) from the local system into a larger environment. If the gate is opened (a “leak” event), some of the accumulated error is removed (like releasing pressure); if it remains closed, the system retains all information for another cycle. The controller must strike a balance:[10] · Excessive leakage: Too many open-gate events will dissipate the system, eroding even useful structure and causing the system to evaporate (analogous to a physical system losing mass or an algorithm discarding valuable state).[12] · Insufficient leakage: Rarely opening the gate leads to error accumulation, a “thermal runaway” where unchecked entropy eventually overwhelms the system.[12] Thus, the Samson V2 strategy is to operate on a knife’s edge, maintaining the system at the edge of chaos – just enough leakage to prevent instability, but not so much as to lose coherence. 1.2 Emergence of a Scale-Invariant Leakage Regime This paper is motivated by an unexpected observation from Phase IV of Nexus simulations. Researchers ran thousands of ensemble simulations of the controller under varying noise conditions. Intuition suggested that a higher background noise (i.e. a larger uncertainty in ) would degrade control performance – one would expect more frequent or more erratic leakage as the system struggles to maintain lock on $\alpha_[13][2][14]$. Instead, two simulation scenarios with 5× difference in noise magnitude produced virtually identical leakage behavior*. In the “low-noise” run, the system leaked with some average probability per step; in the “high-noise” run, that average was the same to within statistical error. The distributions of leak events over time were indistinguishable. This held even as the simulations evolved: the final equilibrium leakage rates were the same in both cases.[15][4][4] Initially, this was suspected to be a bug or artifact. However, subsequent theoretical analysis revealed a deep underlying principle: when the controller’s error estimator variance and the normalization scale (used in gating) increase in tandem, the system enters a symmetry state where the noise scale cancels out. In this Scale-Invariant Leakage Regime (SILR), the controller’s decision metric becomes dimensionless and independent of absolute units, so the probability of leakage depends only on relative error significance rather than error magnitude. This paper provides a full derivation of the SILR theorem and establishes it as a fundamental property of the Nexus control law.[6][16] Beyond the mathematical proof, we will discuss why SILR is important. It represents a form of self-normalization: the controller automatically adjusts to the “temperature” of its environment. If the environment gets noisier, the controller proportionally raises its tolerance, perceiving the world as no less stable in a statistical sense. Conversely, in a quieter environment, the controller naturally tightens its standards. This adaptability is achieved [17]without any external intervention or parameter tuning, purely via the internal structure of the z-score gating. Such scale-invariance is reminiscent of physical laws that look the same across different scales (renormalization group symmetries), hinting that SILR might be tapping into a deeper invariance principle in recursive systems. 1.3 Roadmap of this Paper We begin in Section 2 with a precise formulation of the controller’s mechanism: defining the estimator, the error metrics, and the Z-score leakage gate. This section lays out the equations governing , the leakage probability at time , including the logistic sigmoid activation function that introduces non-linearity into the gating. In [18][19]Section 3, we present a step-by-step derivation of the SILR theorem. We show analytically that under the assumption of a well-calibrated estimator (error distribution matches the reported standard error), the distribution of the normalized error (z-score) is independent of the noise scale. Consequently, the expected leakage rate is proven to be invariant to noise amplitude. This derivation constitutes the core theoretical contribution of the paper.[20][16][21] In Section 4, we validate the theory with simulation results from the Nexus Phase IV simulator. We describe the simulation setup and present metrics from three representative scenarios: (A) low noise, (B) high noise, and (C) a “broken invariance” case where the controller underestimates the noise. The results are summarized in tables and plots, confirming that scenarios A and B exhibit matching leakage statistics (SILR in effect), while scenario C deviates as expected. We include excerpts of the [4][22]verified code used for these simulations – specifically the Samson V2 controller implementation and the configuration of the ensemble runs – to demonstrate the exact logic and to allow reproduction of our findings. All code presented has been executed and tested; we refrain from any speculative pseudocode. Section 5 discusses the implications of SILR for control theory and system design. We explore how a scale-invariant leakage mechanism contributes to robust performance: it essentially provides an internal gauge for error significance, enabling the controller to perform error detection and correction consistently even as external condit