Nested Bubble/Gear (NBG-01) v0.3.0: E2-A Interface-Aware Challenge Preregistration and Pre-Run Freeze

Nested Bubble/Gear (NBG-01) v0.3.0 is the preregistration and pre-run freeze release for Experiment E2-A: Interface-Aware Challenge, the first NBG experiment in this lineage for which predictive closure of the inexpensive observer hierarchy is not guaranteed by construction. E2-A follows the completed NBG-01 control sequence: the v0.1.x positive-control and protocol-repair series established exact closure behavior and a nontrivial trust-separation boundary, while v0.2.0 independently validated the known-fail parity negative control with explicit predictive witnesses. Those controls established that the NBG receipt machinery can both accept a structurally sufficient observer and reject an information-losing observer on controlled dynamics. E2-A changes the transition law by introducing an explicitly frozen interface dependence: q_A = (1 - kappa - delta)m_A + kappa m_B + delta x3 q_B = (1 - kappa - delta)m_B + kappa m_A + delta x4 with p_A = eta + (1 - 2 eta)q_A p_B = eta + (1 - 2 eta)q_B. The frozen parameter grid is: eta = {0.01, 0.03, 0.05, 0.10} kappa = {0.00, 0.05, 0.10, 0.20} delta = {0.00, 0.02, 0.05, 0.10} for a total of 64 preregistered parameter points. The 16 points with delta = 0 form a baseline-continuity slice and must reproduce the corresponding NBG-01 v0.1.2 transition kernels or the challenge run is invalidated before scientific interpretation. The preregistered observer ladder includes the inherited majority observer O0, weight-based observer O1, interface-aware observer O_IF = (m_A, m_B, x3, x4), and the richer observer O2 = (w_A, w_B, x3, x4). Of particular interest, O1 and O_IF have comparable effective description size while retaining different information. This permits E2-A to test whether retaining causally relevant interface information matters more than simply increasing observer detail. Several interoperability rules are frozen before execution following external cross-implementation review of the earlier NBG controls. State serialization is explicitly defined using a canonical MSB-first bit convention. Witness ties are handled deterministically while recognizing multiple maximizing witness pairs as scientifically equivalent. Exact rank deficiency is distinguished from numerical near-rank deficiency, and conditional mixing is evaluated by direct survival-conditioned total-variation contraction rather than by assigning dynamical meaning to floating-point residual eigenvalues. The preregistration also freezes predictive targets, leakage thresholds, composition rules, trust criteria, receipt structure, witness requirements, baseline validation, deterministic export behavior, semantic and implementation digest conventions, and the anti-rescue rule. No observer, parameter, threshold, endpoint, cost rule, witness rule, or mixing convention may be changed after canonical E2-A output becomes visible. No E2-A scientific results are included in this release. The purpose of v0.3.0 is to establish a timestamped, inspectable protocol before execution so that any later result can be evaluated against rules that were fixed in advance. Following publication of this preregistration, the frozen E2-A harness will be executed once, followed by an isolated deterministic replay. The resulting scientific artifacts, receipts, witnesses, and replay comparison are intended for a subsequent results release. NBG-01 v0.3.0 does not establish a new physical law and does not make claims about spacetime, quantum mechanics, consciousness, or nature's literal ontology. It defines a finite, falsifiable computational challenge intended to test predictive closure, observer adequacy, interface relevance, composition, metastable trust separation, and promotion behavior under preregistered conditions.