Gravitational Manipulation Theory v2.4 — Apparatus-Space Scaling of Information Sufficiency

Gravitational Manipulation Theory (GMT) v2.4 extends the information-sufficiency program from a finite three-class robustness test to a pre-registered apparatus-design space. GMT v2.2 established, for a single Controlled Synthetic Apparatus (CSA), operator sufficiency at (L^*{\mathrm{operator}}=\mathcal I_2) and a minimum sufficient record (\mathcal I{\min}=\mathcal I_4). GMT v2.3 showed that these boundaries were retained across three deliberately different apparatus classes, while explicitly leaving open whether that robustness would survive a broader apparatus panel. In v2.4, the apparatus is represented as a point (\mathcal A=(N,\rho,T,O,\mu,\Lambda)) in an apparatus-design space. An eight-class panel, CSA-A through CSA-H, together with the sampling, certification, resolution, and ablation rules, was frozen before any result was observed. The panel spans fold number, radial mismatch, null topology, observable structure, harmonic ratio, differential readout, and a beyond-family linear-grating / translational / spatial-wavenumber apparatus. The range-regime axis (\Lambda) is declared but held fixed in v2.4. One frozen information ablation,[\mathcal I_4\rightarrow\mathcal I_3\rightarrow\mathcal I_2\rightarrow\mathcal I_1\rightarrow\mathcal I_0,]was applied unchanged to all eight classes. The pre-registered mechanical decision returns Outcome A — Full Retention: [L^*{\mathrm{operator}}=\mathcal I_2,\qquad\mathcal I{\min}=\mathcal I_4]for all eight tested classes. The geometry ranges are (G_1=3.13\text{–}6.26) dex, giving (\mathfrak D_H(\mathcal I_1)=3.36\text{–}6.49) dex against the pre-registered tolerance (\varepsilon=0.30) dex. Every class is therefore a certified FAIL at (\mathcal I_1); (D_{\mathrm{op}}=\varnothing) and the indeterminate-(\mathcal I_1) set is empty. The nearest class is the beyond-family CSA-H, with (G_1=3.13) dex, still 2.83 dex above the tolerance. The two retained boundaries arise for different reasons. The operator boundary is geometry-driven, whereas the constraint boundary is disclosure-width-driven:[\mathfrak D_C(\mathcal I_3)=2\log_{10}F_4,]with the exact transition[F_{4,\mathrm{crit}}=10^{\varepsilon/2}\approx1.413.] A provenance audit finds zero hard near-nulls across all eight classes, supporting interpretation of (G_1) as genuine geometry spread rather than accidental cancellation. All apparatus in this work are synthetic representative-class systems; none is HUST hardware. The Yukawa-form kernel is used only as a synthetic response generator. The result supports a class-robust disclosure standard across the bounded, pre-registered apparatus-design space tested here, not a universal law. No experimentally demonstrated gravitational manipulation is claimed. The accompanying supplement contains the frozen Master Specification, proposal, Stage 1 validation and Stage 2 result reports, source code, numerical JSON outputs, and Figures 1–7. The natural next step for GMT v2.5 is to span the range-regime axis (\Lambda) and further widen the beyond-family apparatus panel.