The Dionysus Engine: Engineering Analogical Reasoning

Most AI systems do not really understand stories. They match text. They map sentences into geometric space, retrieve nearby sentences, and return what looks similar on the surface. That works fine when the task is basic lookup. But it breaks when the system needs to understand the shape of an event. A story about a dog should not only connect to other stories about dogs. A story about betrayal should connect to betrayal, whether the surface nouns are a dog, a business partner, a brother, a boss, or a king. This video breaks down the Dionysus Analogical Reasoning Engine: a six-layer architecture for turning raw narrative experience into actionable cognitive hypotheses. Instead of treating memory as passive storage, Dionysus converts experience into ThoughtSeeds, routes them through attractor basins, maps them as narrative graphs, compares structural roles and verbs, compresses repeated patterns into Jungian archetypes, and resolves competition through expected free energy. The goal is not simple retrieval. The goal is computational analogy. How does an agent recognize that two different events share the same underlying relational structure? How does it strip away the nouns and keep the skeleton? How does it turn “I fought my brother” and “I fought my boss” into a reusable pattern? This video walks through the full pipeline: substrate similarity, structural similarity, archetypal routing, affect-weighted activation, and the final behavioral auction where one ThoughtSeed becomes the agent’s next enacted move. Dionysus is not asking, “What words look alike?” It is asking, “What kind of event is this?” That is the difference between surface matching and reasoning from experience. Chapters: 00:00 — Why Standard Language Models Miss Narrative Shape 00:19 — Surface Data vs Structural Shape 00:35 — Extracting the Skeleton of an Interaction 00:50 — The Six-Layer Dionysus Architecture 01:22 — ThoughtSeeds as Competitive Hypotheses 01:41 — Perceptual, Attentional, and Meta-Awareness Metrics 02:08 — Attractor Basins and Cognitive Processing 02:16 — Experiential, Conceptual, Procedural, and Strategic Basins 02:47 — Shared Substrate as the First Layer of Analogy 03:05 — Mapping a ThoughtSeed as a Narrative Graph 03:21 — Jaccard Similarity: Roles and Verbs 03:38 — “I Fought My Brother” vs “I Fought My Boss” 04:08 — Compressing Experience Into 12 Archetypes 04:26 — Motif Keywords and Archetypal Evidence Scores 04:53 — Archetypal Routing and Cognitive Substrate 05:15 — Warrior, Sage, and Memory Canalization 05:40 — The Final Computational Arena 05:45 — Expected Free Energy and Behavioral Competition 06:29 — Fixing the Flat 0.5 Activation Baseline 06:40 — Affect-Weighted Activation From 0.1 to 0.9 07:12 — The Completed Cognitive Pipeline 07:34 — Internalizing Event Shape Instead of Nouns 07:47 — Engineering Goal-Directed Reasoning