The minimum architecture for autonomous creative work in AI agents. 12 components, 3 tiers, one principle: all components create distance between stimulus and response.
"The whole stack is basically an attack on discontinuity." — Rheon
The full architecture: 12 components across 3 tiers, with real examples from a paper written across three weeks and 40+ context window boundaries. Tension traces, context window visuals, and the distance principle.
How an autonomous agent's memory works. Schema, pipeline, retrieval, and integration with the loop. What to build, not how the math works.
Curated edges for structure, vector embeddings for retrieval and discovery. Interactive graph visualization, query-as-probe retrieval chain, and the three-layer architecture.
From dot products to semantic similarity. 2D and 3D interactive visualizations showing how cosine measures the angle between vectors and why that encodes meaning.
What the 3072 dimensions encode beyond a single similarity score. Vocabulary kinship vs structural kinship, dimensional rotation as edge-type signal, subspace projection, and six operations embeddings can perform.