Built, and designed
A reference site for a primitive that five other sites depend on owes its readers this table before it owes them anything else. Every claim made anywhere on this site appears here with its state, and the states are not softened.
1. The table
| Artefact | State | Evidence, or the absence of it |
|---|---|---|
| The S3-compatible vault container | built | Working. "most of the code exists already", and the sg-compute analysis confirms Routes__Vault__Spec and the vault specs ship. Swappable vault, disk and memory backends behind one S3 face |
Twin in the grounding ladder | primitive | Established and cited across the estate: Measure := an observation of the node it measures, grounded on a Twin |
connected_to in the edge grammar | primitive | Established. The established edge set ends with the connection from any node to a twin and onward to reality |
| The AWS twins in the IAM config risk engine | partial | Spec with real code behind it: Python, unit-tested, JSON out. Reuses "the digital twins already built", which the site cannot currently point at |
| The coverage measure | specified, never run | Fully specified over the regulation graph. Has never been computed. One number would make the discipline page concrete |
| The execution broker | designed | A design, converted from an external session. Two costs published: concentration risk, and one interpreter per provider per capability |
| Twins as actors, personas, the change pipeline | designed | Nothing built. The 2FA demo is the milestone that would evidence it |
| The 2FA demo | designed | Not built. Everything downstream labels itself against whether it exists |
| The world-model simulation | designed | Not built, and in unresolved tension with the discipline of reality |
| The organisation twin and the agent twin | designed | Not built. The agent-twin brief carries its own objection about how connected a twin really is |
| A typed twin schema | absent | "properties, behaviours, functions, inputs and outputs" is stated everywhere and typed nowhere |
| The naming ruling | this site | Built fresh here, because the corpus flagged the collision and left it unresolved. Enforced in CI by the name-watch |
| The name-watch check | built | Runs on every release in admin/build/validate.js. A third sense of the word fails the build |
Two built, two established, one partial, one specified-never-run, one absent, six designed. The ratio is the finding, not an embarrassment: it is what a corpus looks like when a primitive gets adopted faster than it gets implemented. What it costs is that one container is carrying the entire interface claim on its own.
2. The corpus behind it
These counts were measured on 24 August 2026 by the librarian who assembled the brief pack this site is built from, and they date rather than refresh.
148 files
mention "digital twin"; 247 mention twins at all. Eleven twin-named briefs, roughly 28,000 words of dedicated treatment.
26 June 2026
Four briefs, 6,863 words, written in one cluster: twin-of-anything, the integration layer, the 2FA bottom-up build, and world models. Most of this site comes from here.
June to August
Identity visualisation (4 Jun), the S3 twin (23 Jun), the 2FA capstone (26 Jun), the AWS engine (5 Jul), hyperscaler abstraction (23 Jul), twins-as-hooks (28 Jul), the broker and its market research (19 Aug, 8,488 w).
8 documents
A manifest of 22 rows, every path verified on disk. Published here verbatim under CC BY 4.0. Read them →
3. Where the material is weakest
Worth saying plainly, because the pattern is not random: the best ideas arrived last and are the least verified. The 19 August pair is 8,488 words, converted from a voice session with a different assistant, with a market-research companion marked "as supplied". The broker is the strongest single idea in the corpus and carries the weakest provenance in it. This site quotes the brief with the note attached and quotes the market research not at all.
The five honest tensions, including that one, are set out in full on the open questions page.