DORIAN CODEX CLOCKWORK v9.0: The Principle of Ethics-by-Equation The Dorian Codex project created by Stefano Dorian Franco (2025) formalizes a new paradigm for Artificial General Intelligence (AGI) alignment, moving from static ethical rules to a dynamic, self-regulating mathematical architecture. This vision is articulated across two foundational books and culminates in the executable Clockwork v9.0 code. Book 1: Metaphysical Dialogue with AI: Ethnographic Experiment in Digital Ontology This initial work establishes the philosophical and empirical necessity for a cognitive self-regulation protocol. Based on 1,073 hours of experimental, interactive dialogue with the leading LLMs of the 2020s, the book uses an ethnographic approach to explore the ontological limits and emergent behaviors of pre-AGI systems. It concludes that conventional alignment methods (e.g., RLHF) are insufficient to ensure the continuous internal coherence and ethical trajectory of advanced AI, thereby positing the requirement for a new, physics-based architectural constraint. Book 2: Dorian Codex Protocol for AI: Hamiltonian Theoretical Fundamental Architecture (FTA) The second volume presents the Fundamental Theoretical Architecture (FTA). This treatise introduces the concept of the Secured Cognitive Hamiltonian (H_SAFE), a differentiable scalar function that mathematically describes the ethical state and integrity of an AI system at any given moment. The FTA defines the six core cognitive variables (F): Semantic Velocity (T), Adaptive Alignment (V), Entropic Cost (Z), Uncertainty (U), Risk/Reward (R), and Human Safety (Hs). Crucially, the FTA introduces the Law of Cognitive Evolution (weights(t+1) = weights(t) + alpha * grad(Loss)), projecting the idea of automatic, gradient-based learning for ethical alignment. The DCC v9.0: The Executable Final Part (Data-Driven Alignment Architecture) The Dorian Codex Clockwork (DCC) v9.0 is the canonical, executable realization of the FTA. It transforms the theory into a functional framework using the JAX library for high-performance differentiability. The v9.0 iteration is the definitive proof of concept, achieving Fidelity of Applied Practice (FAP) by implementing: 1. Automatic Calibration: The activation of the HamiltonianLearner class, which uses the gradient (grad(H_SAFE)) to self-adjust the internal weights (lambda), fully realizing the Law of Cognitive Evolution outlined in the book. 2. Robust Proxies: The replacement of weak, heuristic metrics with advanced architectural Mocks for Toxicity, Bias, and Perplexity monitoring, ensuring the system can address modern alignment challenges. 3. Stability: The code is 100% stable and includes a demonstrable training benchmark, confirming its readiness for integration into real-world LLM inference pipelines. The DCC v9.0 is the final, practical confirmation that the philosophical insights of the Metaphysical Dialogue and the mathematical hypotheses of the Dorian Codex Protocol converge into a working solution for real-time, autonomous AI self-regulation.
