The Throughput Basin: Cross-Substrate Convergence and Decomposition of Serial Decoding Throughput

Building on the rate-distortion floor established in a companion paper, we examine throughput convergence across six domains: genetics, artificial intelligence, human cognition, music, language, and engineered encodings. Across 31 systems with vocabulary sizes spanning 64,000×, effective information per serial event occupies a 3–6 bit band. We propose the Throughput Basin Decomposition: I_eff = R_M(ε) + Δ_s + ξ. Three independent evolutionary simulations converge to K ≈ 19–30 and ~4.1–4.6 bits. A permutation test reveals that cross-substrate statistical significance depends on AI data inclusion — a critical caveat we report transparently.