The Functional Threshold of Intelligence: Why No System Is Intelligent Until All Core Functions Are Implemented

This paper introduces a minimal functional model of intelligence defined not by outputs, performance benchmarks, or training regimes, but by the recursive operability of six internal functions within a semantically grounded functional state space. We call this framework Human-Centric Functional Modeling (HCFM). It posits that any system lacking even one of these functions—such as recursive modularization, functional reconfiguration, or cross-domain generalization—is not partially intelligent but structurally unintelligent. The model is self-validating: we demonstrate, through recursive co-authorship with a state-of-the-art language model, that current AI systems fail precisely in the ways predicted by HCFM. These failures—semantic drift, conceptual collapse, visual incoherence—serve as functional diagnostics for detecting the absence of true recursive intelligence. To scaffold implementation, we decompose the six core functions into five architectural capacities and present a minimal engineering substrate. However, this paper itself is ultimately scaffolding. Once internalized, the model can be compressed into a short animated visualization, understandable without mathematical formalisms, yet infinitely generative in its capacity to produce them. Intelligence, as defined here, becomes accessible not through accumulation of knowledge, but through recursive reorganization of function. This work aims to realign the foundation of AI development, safety, and epistemology around that principle.