The simulation layer
designed Nothing on this page is built. It is the most ambitious material in the corpus and the least evidenced, and it is published as a roadmap rather than as a capability, with one concrete milestone at the front of it.
1. The idea
If a twin is a system with properties, behaviours, functions, inputs and outputs, then a collection of twins is a world. The 26 June strategy brief takes that literally: assemble twins into a simulated environment, "an org chart, actions, risks taken, compromises, and clients buying, built from mini agents", and run a company inside it.
The vault is the engine, and the reason given is a systems reason rather than an aesthetic one: "every action is version controlled and every connection is mapped by folders and files without a database ... a showcase of serverless databases." The history of the world is the history of the vault. There is no separate simulation state to keep consistent, because the state is the commit graph.
2. Why game design is the explicit analogy
Not decoration. The observation in the source is practical: "LLMs are good at generating worlds and games", so the architecture that game designers already use for agent-populated worlds is the architecture to borrow. Mini agents with local goals, a tick, a world state, an event log. MITRE ATT&CK layers on top as the adversary's move set.
3. The tension this page cannot resolve
The corpus's most distinctive rule and its most ambitious design pull in opposite directions, and this is where they meet.
| The discipline of reality | The world model | |
|---|---|---|
| Admits | Only what exists, with evidence | Attacks that have not happened, clients who have not bought |
| Its value comes from | Refusing the hypothetical | Exploring the hypothetical |
| Its failure mode | Covers less than you hoped | Pollution: exactly what the discipline exists to prevent |
The likely answer, and it is only likely: simulated states are marked as projections from real twins, never as facts, and never enter the graph as evidence. A projection would carry the twin it was projected from and could not be cited by a measure. Nobody has written that rule, and until somebody does, a world model built on this graph is a way to pollute it. Question 3 →
4. Milestone one: the 2FA demo
The roadmap's first concrete step is not the world. It is one bottom-up build: organisation, HR system, people and roles as twins stored in IssueFS, assembled from the bottom rather than declared from the top, with abstraction layers placed at the real seams, and a risk acceptance run through the change pipeline end to end.
The reasoning for starting there is stated in the source and is worth keeping: "develop the core pieces of technology on top of these use cases, which are both a good way to build the technology and a good way to explain how it works." A demo that has to work is a specification that cannot be hand-waved.
It is gap G4. Everything on this page is downstream of it.
5. The honest warning that comes with it
The same brief that recommends starting small says why: "start with one simple mapping and see how big it gets, because even one specific use case can become massive once you capture all the elements."
That is a warning about scope, and it contains an unanswered question. If capturing one use case fully can become massive, something has to say when a twin is finished. The grounding ladder's stopping test, "the last node where going deeper would neither improve observability nor change a decision", probably applies here too. The corpus does not say so. Question 5 →