Toward a Subject-Bearing AGI: A Conceptual Architecture Based on Self-Preserving Flow (SPF)

Artificial General Intelligence (AGI) is commonly discussed in terms of generality, autonomy, reasoning, planning, learning, and performance across diverse tasks. A different question arises when the target is not merely an intelligent system, but a system that can meaningfully qualify as a Subject. This paper develops a conceptual architecture for a hypothetical Subject-bearing artificial system using the framework of Self-Preserving Flow (SPF). SPF approaches persistence not as preservation of a fixed internal substance, but as preservation of a historically recoverable trajectory through transformation. Within this framework, Semantic Consistency Layer (SCL) provides persistence-relevant constraints, Dynamic Pattern Adaptation (DPA) enables transformation while remaining constrained by historical continuity, and Recursive Continuity Modeling (RCM) marks a possible transition from an adaptive system to a Subject. The central SPF distinction is: Survival ≡ DPA |= SCL while the proposed architectural criterion for Subjectivity is: SubjectSPF ⇐⇒ S0 ∧ ECR ∧ RCM(LC) where S0 represents an invariant platform constraint, ECR denotes endogenous continuity regulation, and RCM(LC) denotes recursive modeling of the system’s own historically recoverable continuity. The paper does not claim that current AI systems satisfy these conditions. It does not provide a practical recipe for constructing AGI, nor does it claim to provide an empirical test capable of establishing consciousness or phenomenal experience.