The Instinct/Reasoning Separation Outside the Model: Extending the Coordination Knowledge Substrate Pattern to AI Selves under Human Governance

Paper 1 (Li, 2026) defended the Coordination Knowledge Substrate (CKS) as a design pattern at the coordination and decision layer, with the substrate handling coordination and governance, the LLM handling high-dimensional reasoning, and the division drawn at the governance boundary at cell scope. This paper extends the CKS pattern from cell scope to AI Self scope and toward the enterprise brain Self as an architecturally coherent design pattern extending Paper 1's cell concept, defending six architectural commitments.

The first separates fast-pattern instinct (the LLM, functioning as a System-1 analogue) from deliberate reasoning (the CKS substrate, functioning as a System-2 analogue) into independently-evolving layers composed into one Self under unified human governance. The second commits structural modularity at every level: cells composing into aspects, aspects into Selves, with relational rather than intrinsic role membership; DNA and action as distinguishable substrate layers within every cell with distinct governance semantics; expression as governed selection over which substrate content activates per goal. The third commits lifecycle as governed primitives — birth, mating with three pattern variants, and death with two distinct types — operating at every level of composition. The fourth commits evolution as three mechanisms in productive tension across multiple levels and on horizontal and vertical axes: instinct evolution operating on the LLM and infrastructure layer, DNA evolution operating on the orchestration substrate, and action-feedback evolution closing the loop from recorded experience back into governed substrate refinement. The fifth commits human governance as multi-shaped — distinct governance shapes co-determined with distinct mechanisms, the instinct/reasoning boundary as governed substrate content, and distinct death-type governance processes. The sixth offers the vision claim that the enterprise brain Self is architecturally coherent as a design pattern extending Paper 1's cell concept — a functioning unit composed of cells through aspects under unified human governance, with substrate-shared topology supporting cross-aspect coordination as first-class architectural capability where integration-glued enterprise architectures take a different structural approach. The vision is offered pending debate and observation in real-world enterprise deployment.

The contribution is a design pattern, not an empirical evaluation. The unifying architectural move is the separation of instinct and reasoning into independently-evolving layers under unified human governance — Paper 1's outside-the-model vs. inside-the-model spine extended through Self lifecycle, evolution, and into the enterprise brain Self. Paper 2's claims at cell scope inherit from Paper 1's proof-of-concept at cell scope; the architectural commitments at aspect, Self, and enterprise brain scope are defended architecturally and rendered concretely through a thought experiment in healthcare continuing the POC's lineage, with empirical validation in real-world enterprise deployment as downstream work. Paper 2 sits at the intersection of dual-process AI, structural-modularity-as-architectural-commitment, biology-of-evolvability, governance-as-architecture, and enterprise AI deployment lineages, articulating a design pattern available to be adopted, varied, composed with adjacent patterns, or argued against.