Inside · unlisted
The engine, named from the inside.
The front of this site describes one engine that stands up self-running worlds. This page is the same engine with its internal names on. You do not need any of it to use the thing.
Back to how the engine works →These names came out of building it, not out of marketing it. They are kept here because they are load-bearing in the code and in the repos — and kept here, unlisted, because a stranger meeting them at the door learns nothing.
The stack
Seven layers, one engine.
Each layer does one job and makes the layer above it cheaper. Read top to bottom: the human, then what the human knows, then what acts on it, then where it acts, then how it is watched, then how it rebuilds itself, then how what it learns gets out.
The life layer
Health, wealth, relationships, purpose, growth, environment — tracked and connected to everything else. The layer that keeps the human intact while the rest of it runs.
The memory layer
Insights, patterns and decisions persisted as typed relationships rather than as files. Semantic search finds what you forgot; the edges surface what you never connected.
The agent layer
A personal infrastructure agent that reads the graph, builds skills, writes code and runs the back office. Not a chatbot — a process with a memory that outlives the session.
The business layer
Departments, SOPs, automations, dashboards. What a company looks like when its org chart is a running program. This is the layer sold as Run.
The runtime layer
The agent runtime everything above executes on, with ontological validation and logic enforcement in the loop — the work is checked against the rules before it acts, not after.
The development layer
Every codebase is a starsystem: flights, missions, starlog, waypoints. How the system builds itself, records what it did, and teaches the next pass to build on top of it.
The distribution layer
Knowledge-graph concepts become notes. Notes become diagrams. Diagrams become animated explainers. Explainers become video. That is why the marketing is not separate from the product — the product produces the marketing.
What this page is not
Names are not evidence.
A layer having a name does not mean it is finished, and this page deliberately carries no counts — no line totals, no concept totals, no "N frameworks published". Those numbers were on the old version of this page, none of them were checkable from outside, and one of them contradicted the number of posts actually on the site.
The claims worth anything are the ones you can open yourself. They are on the map and in the notes, and the code is public.
Receipts, not testimonials
The parts you can read.
Two of these layers are public repositories. Open them before believing anything above.
The monorepo
The engine described on this page, with its commit history.
github.com/sancovp →The extracted tools
heaven-framework, chaincompiler, carton-mcp, skilltree — MIT.
All repos →Layer 07, live
The content pipeline drawing its own explainer in your browser.
Watch it →How each layer was built
The notes were written while the layers were being built, mistakes included.
Read them →Also inside
Why it is built this way
The ethics-as-architecture argument, and the one part of it that is a real mechanism rather than a metaphor.
Read →GNOSYS
The agent layer, described on its own terms.
Read →The Sanctuary System
The myth that was written before the code, kept for the record.
Read →Completely Fake Dharma
Unedited. The source material the ethics argument came out of.
Read →