AI system that combines reinforcement learning (RL), structured knowledge graphs, and neural retrieval techniques to dynamically select and reason over specific facts and concepts or training data. It enable even small LLMs to produce clearly structured, understandable detailed Chain-of-thought, active learning.