Data and code for: Cross-Domain Evidence for Bottleneck-Record Systems - Structural Persistence, Validity Decay, and the Limits of Lossy Compression

This deposit contains the data and code supporting the PLOS ONE submission "Cross-Domain Evidence for Bottleneck-Record Systems: Structural Persistence, Validity Decay, and the Limits of Lossy Compression" (PA-P030). The paper reports cross-domain directional evidence for a Bottleneck-Record construct across four empirical domains (DSM diagnostic classification, US Code of Federal Regulations growth, US urban scaling, and computational autoencoder experiments), together with a honest null boundary result under lossy compression. The deposit includes: (i) four computational experiments (SIT-1 bottleneck-bloom structural inheritance; SIT-2 lossy compression with both fixed-codec and learned-codec arms; SIT-5 distribution-matched control extension; and the H-004 extended-widths post-hoc re-run) as self-contained Python scripts, run configuration YAMLs, raw result JSONs, per-run metrics TSVs, and comparison-table TSVs; (ii) figure-generation scripts producing all six main-body figures and the three supplementary figures from documented source values; (iii) the five PA-P030 supplementary tables (full evidence registry, pattern canon references, contradiction references, CAN-018 directional evidence, and CVE retrospective support); and (iv) the four newly deposited evidence rows (EVI-020..023) together with a knowledge-graph reconciliation note. Important scope caveats are preserved here and in the bundle's top-level README: the formal H-001 CVE prediction test was never run (NVD API access was blocked in the hardening lane) and is not included as completed evidence; the H-002 DSM category replication remains non-significant and is not silently upgraded; the SIT-2 result remains a confirmed null / regime-boundary under both codec families; and the DSM logistic fit (EVI-023) is a quantitative fit, not a standalone causal proof.