Abstract This dataset contains the theoretical proofs and experimental verification for the Inverse Scaling Law (ISL). We propose that thermodynamic constraints impose a hard physical barrier on non-modular computation. Included Materials: ISL_Framework_Paper.pdf: Unified theoretical paper (Typeset). supplementary_proofs/: Detailed Theorem Proofs (PDFs). Theorem_01_Modularity.pdf Theorem_04_Reuse.pdf Theorem_07_Scope.pdf code/kill_switch_experiment.py: Source code for the generalization gap experiment. figures/: High-resolution verification plots. Key Results: Experimental validation confirmed a 38x efficiency gap between modular and monolithic architectures in data-starved regimes, supporting the Information Reuse Bound ($T \propto 1/R$). References: Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM Journal of Research and Development. Shannon, C. E. (1948). A Mathematical Theory of Communication. Bell System Technical Journal. Gromov, M. (1987). Hyperbolic groups. Essays in Group Theory. Bridson, M. R., & Haefliger, A. (1999). Metric spaces of non-positive curvature. Springer. Kolmogorov, A. N. (1965). Three approaches to the quantitative definition of information. Problems of Information Transmission. Bekenstein, J. D. (1981). Universal upper bound on the entropy-to-energy ratio for bounded systems. Physical Review D.