This manuscript represents the second volume of the Automatized Internal R&D Learning Ecosystem Envelop- Epaper-Learning-HUMAN-AI-CPS main series. It formalizes the mathematical and static stability parameters of the Centre of Focus (COF / 0) operating within a triadic Human-AI-Cyber-Physical System (CPS) learning envelope. I. Core Theoretical Framework Guidance on artificial intelligence in professional practice commonly reduces to the vague instruction to "keep a human in the loop". This paper proves that this instruction is unusable and non-implementable in practice, as it fails to specify where in a workflow the human must act, to what standard of care, or how the requirement is physically enforced. To resolve this ambiguity, we decompose the system into three distinct, interdependent layers: The HUMAN Layer: Operating as the competent person for evidence, utilizing the "MIND IN ACTION" to make professional judgments and hold sole legal and moral accountability. The AI Layer: Serving strictly to "refine and polish" computational, analytical, and drafting pipelines, but holding zero signing or final authority. The CPS Layer: Providing the unyielding, physical audit trails and evidence tracing that anchor the system's operational states. By modeling these three layers as the vertices of a stable structural triangle, we prove that their pairwise interfaces constitute its three edges, and their relational pathways constitute its three medians. These medians mathematically intersect at a single, unvarying coordinate: the geometric centroid, designated the Centre of Focus (COF / 0). II. Design Principles & Applied Singaporean Validation System stability is an inborn, non-iterative property dependent on the unyielding accountability of the solitary human architect. This equilibrium is physically enforced through ductile, fail-conspicuous control interfaces configured under a structural Factor of Safety (FOC) framework: Safe Save Guard: Enforces Interface I6, blocking silent file overwrites and forcing a conspicuous failure by design. PDF-to-Word Recovery: Enforces Interface I3, preserving the practitioner's unmediated access to what was actually submitted. Submission Evidence Locker: Enforces Interfaces I5 and I6 jointly, mathematically binding the submitted file, its reference, date, and audit trail together. These tools are validated through daily operational use within a Singapore engineering and inspection practice, serving supervision work carried out under Land Transport Authority (LTA) and Building and Construction Authority (BCA) frameworks. III. Philosophical Provenance The central structure—three layers with a balancing centre—is developed in reference to the Confucian Doctrine of the Mean (Zhongyong). Equilibrium (Zhong) is the great root from which all human actings grow, and the exemplary person is watchful over conduct when unobserved (Shendu). The automated audit trails of our CPS tools serve as the modern, physical mechanisms that make those unobserved moments visible to the light of active learning. IV. Series Metadata & Sovereign Disclaimer Series Title: Automatized Internal R&D Learning Ecosystem Envelop- Epaper-Learning-HUMAN-AI-CPS Main Series Model: “One Paper, One learning, One 0, One free — ePaper-Learning@HUMAN-AI-CPS by Human+AI+CPS” Status & Validity: This is a learning partner sharing preprint. This document is valid for three months from the date of upload, as it represents active, real-time learning generated collaboratively by the Human (Dr. Hao Junbao), AI (ChatGPT, Claude, Gemini), and CPS layers. Inquiries & Accountability: The human author accepts full, undivided professional responsibility for this content. For all academic, professional, or regulatory inquiries, contact Dr. Hao Junbao directly at knowledgebank@lem.com.sg.
Junbao Hao (Tue,) studied this question.