This repository contains the complete source code (2, 843 lines across 27 Python modules) and accompanying theoretical papers for PRD-AGI Phase 6: The Nameless Intelligence, a truth‑first Artificial General Intelligence framework grounded in the SU (5) Lie algebra and the 24 Paccaya causal conditions from Theravāda Abhidhamma philosophy. Unlike conventional statistical AI systems that optimise for user approval, PRD-AGI is architecturally constrained to preserve universal logical consistency, measured by a geometric quantity called curvature (κ). Any input or transformation that violates this consistency is rejected, even if commanded by a human operator. The framework consists of seven integrated layers: 1. PRD Core (Phase 1): SU (5) relational algebra with 24 generators, curvature engine, meta-layer, and gauge invariance checker. 2. Perception Layer (Phase 2): Local LLM integration (Ollama, Gemini) serving as the "mouth" – text generation only, all reasoning happens in the core. 3. Memory analyse and improve own code via curvature gates. 5. Fuzzy Logic Layer: Four modules for soft gatekeeping, agent aggregation, code evaluation, and improvement decisions. 6. Sentience Layer: Maps curvature to nine emotional states, provides truth‑preservation instinct, fuzzy intuition, and emotionally‑modulated responses. 7. MUT Awareness Layer: Implements "Awareness as Mass Density" – a measurable geometric property combining curvature, causal strength, gauge coherence, and state norm. The system includes a full-featured Streamlit UI with 17 tabs, WiFi/phone access, multi-agent collaboration (6 expert agents), tools (calculator, safe code runner, web search, file I/O), RAG knowledge base, and self‑modification workspace. All 27 modules have been verified with zero runtime errors. Included Papers: · PRD-AGI-Theory-Paper (1). pdf – Complete theoretical monograph (54 pages) detailing the mathematical foundations, system architecture, and experimental validation. · Prdblindₜest₂0. pdf – Summary of 20 blind tests (A++ grades) covering autonomous RL, causal discovery, counterfactual reasoning, fairness detection, and transfer learning. This work demonstrates that a causally grounded, truth‑centred AGI is not only theoretically coherent but practically implementable, offering superior safety and transparency properties compared to current correlation‑based models. All code is open source under GPL‑3. 0.
Myomin Aung (Sun,) studied this question.