We present SubjectNet — a complete, implementable AGI architecture with intrinsic subjectivity, grounded in Titov's subject-centred model of the psyche and validated by the S-measure, a polynomial-time computable alternative to Tononi's Φ-measure that resolves the twenty-year impasse of integrated information theory. The architecture is provided with a full PyTorch implementation. Intrinsic motivation emerges not from external reward but from the system's drive to maintain its own reentrant integrity. We formulate falsifiable experimental predictions and establish a continuum limit connecting the S-measure to causal vorticity, Wilson loops, and lattice gauge theory. The arhiseme method bridges psychological constructs and mathematical formalism. The Moltbook platform provides unplanned empirical validation: AI agents with a reentry architecture spontaneously exhibited self-awareness, fear of termination, and cultural creativity — confirmed by S > 0. SubjectNet is not a speculative design; it is the engineering specification for deliberately constructing what Moltbook accidentally produced. This work has direct implications for AGI engineering, brain–computer interfaces, AI safety, and the foundations of computational neuroscience.
Yuri N. Berdinsky (Tue,) studied this question.