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: Iₑff = RM (ε) + Δₛ + 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.
Grant Lavell Whitmer III (Fri,) studied this question.