World-Building System: Generative AI, Knowledge Graphs, and Rules
Three live demos show how rules, memory, and graph templates combine into a coherent world-building system for AI-driven stories and visuals.
Projects
Recent Posts
Introducing GrooveGraph: One Graph for Every Facet of Recorded Music
GrooveGraph unifies recorded-music credits, sessions, and provenance into one graph so discovery becomes traversable and curatorial work stays traceable.
Introducing Alexander: Describe a Place, Get a Pattern-Grounded Vision
Alexander turns a place description into pattern-grounded narrative and image concepts using Christopher Alexanders patterns as evidence.
StoryWorld: Chat With a World, Not Just a Model
StoryWorld gives a story universe memory, graph-backed structure, and guardrails so chat can grow a coherent narrative world instead of a one-off prompt.
The Research Loop Leaves the Lab
There is a difference between saying a future product will be discovery-first and watching a real session produce sources, evidence, claims, and graph-shaped candidates while you are still deciding what the ontology ought to become.
The Janitor Gets a Badge
There is a stage in the life of a technical system when everyone still says the word "cleanup" in the tone of a civic virtue and nobody can point to the broom.
GG-08 - 2026-03-18 - Hardening the Chat-First Surface
Hardening the chat-first surface means removing leftover logic, simplifying interaction, and treating render churn and UI feedback as first-class bugs.
GG-07 - Query Builder Near-Complete Wrap (2am)
Ontology-invalid proposals become actionable as users accept graph changes directly and the system persists them to Neo4j with traceable decisions.
GG-06 - 2026-03-17 - The Day We Stopped Guessing
The chat-first pipeline stops guessing by keeping the LLM central and replacing stale builder-era assumptions with explicit decisions and diagnostics.
GG-05 - Guided Next Rows and Direction-Aware Chains
Suggested next connections and per-row direction make the builder follow graph reality instead of forcing users to translate guidance manually.
GG-04 - Fuzzy Interpretation: First Live Loop
The first fuzzy interpretation loop blends prior insights, model-guided interpretation, and ontology heuristics to hydrate live query state.
GG-03 - Multi-row Composition, Still This Afternoon
Multi-row composition becomes real as chained intent compiles live and invalid relationships fail fast while valid chains succeed.
GG-02 - Live Query Builder Slice
A live query-builder slice sends one row straight to the compile API and keeps the full response visible for debugging.
GG-01 - Fuzzy Foundation, This Afternoon
A foundation-first refactor split the repo into frontend, backend, and utilities while choosing fuzzy, evidence-driven behavior over premature certainty.