Harness-Centric Intelligence

The brief's counter to model-centric takeoff stories: capability in deployment is a property of the harness — tools, memory, shared knowledge bases, agent protocols, institutional interfaces — not of the weights alone. The same weights in a richer harness outperform a stronger model alone, and the harness improves "on a different clock, driven by usage rather than training." Hence "the unit that learns is the ensemble, not the agent," and what it learns accrues to the commons between agents. Generalized in §10.6 as AHI (artificial hybrid intelligence): intelligence explosions are substrate-federations (many systems coupling) rather than substrate-consolidations (one system scaling), so the monolithic Singularity is called "a category error." Symbiogenesis applied to intelligence; cf. WII's "intelligence is an ecology."