The Three Limits of AI Acceleration: Industrial Materialization, Carbon–Silicon Cognitive Asymmetry, and the Causal Horizon of Human Supervision

Contemporary debates about slowing frontier artificial intelligence are predominantly framed as a contest between capability acceleration and safety capacity. This paper argues that this representation conflates three structurally distinct limits that can independently constrain sustainable AI acceleration. The first is an industrial materialization limit: algorithmic capability can advance faster than semiconductor fabrication, memory production, energy systems, data-center construction, networking, capital deployment, and supply chains can materialize it at civilizational scale. The second is a carbon–silicon cognitive limit: biological cognition may be unable to observe, assimilate, contradict, and reorganize itself at the velocity at which algorithmic systems transform an increasingly coupled human–AI environment. The third is a causal-observability limit: even a cognitively capable human supervisor may be unable to exercise meaningful oversight when the causal genealogy relevant to the trajectory being supervised is no longer sufficiently recoverable. The paper further proposes the Supervisory Recursion Principle (SRP): a supervisory intervention can modify the causal object that subsequent supervision must reconstruct. Human participation therefore does not necessarily produce a monotonic increase in effective human control. The framework distinguishes structural limits from the attractor fields through which societies respond to them and proposes an experimentally discriminable distinction between cognitive limitation and causal-observability limitation. The resulting framework shifts the pacing question from whether AI is simply advancing “too fast” toward three independent questions: what civilization can materially build, what biological cognition can adequately follow, and whether the causal history required for intervention remains recoverable from the jurisdiction that must act.