What is shipped, what is argued
Nothing on this site is running yet. The articles are real and dated; the newsroom is a design. Here is the line between them. Per the project's own reality tracker, the entire evidence-economy cluster this site describes is marked PROPOSED — does not exist yet.
What runs
The vault substrate
Articles-as-vaults, mini-site deployment, vault CI — this belongs to sgit.ai, which builds and runs it. Cited here, not re-explained. The boundary →
Ten published articles
Real, dated, and independently checkable: the library, published on docs.diniscruz.ai between February and October 2025.
MyFeeds as a live pipeline
A personalised-news pipeline exists and operates. What it does and how it is priced cannot be published in detail on this site — see comms for why. Its eight-phase rebuild is a design and did not happen.
The governance graph and its gate
Newly real, and narrowly: the typed graph, its published vocabulary, the source register and the build gate that rejects an unknown edge verb all run. The ingestion and verification layer does not — nothing is frozen, hashed or anchored, so no claim there is a checked reading of any text. The limits →
The newsroom, and the door that is shut
Also newly real: the floor, whose seven desks answer from live data rather than a script, and the state map, which names for every state the door it has to get through. One door — frozen, a hashed byte copy of the source — has never been passed by anything, which is the whole reason every fact in the section is secondary. The Researcher's runs are published, including the two primary sources our own egress blocked.
The x402 payment rail
Real, operating, third-party infrastructure — verifiable independently of this site. The rail →. No sgit.ai integration with it exists.
Designed only
Every newsroom brief, every evidence-pack brief, and specifically: Trust-as-a-Service, author micropayments, contextual validation as a billable unit, CC-Signed as an enforceable licence, the decision graph, supersede-never-delete and typed citation edges, the fifteen newsroom departments, the eleven-to-fourteen agent roles and the daily operational clock, the £8.40 worked story ledger. All of it is specified in the project's own briefs. None of it has been built.
And every instance in /mvps/, with one partial exception. The Governance Wire now has a first MVP whose structure genuinely runs; what it does not have is the ingestion path that would let it reach a primary source, so it publishes nothing a reader should rely on and says so on every page.
That section is the closest this site comes to a launch plan — a named country publication, a dated deadline, a phased ramp, ten acceptance criteria — which is exactly why it belongs on this list rather than reading as an exception to it. No domain was registered, no story was published, no agent role was run. The targets recorded there (a ≥200-entity graph, “the first thirty days”, five articles before launch) are goals that were set, not results that were reached, and a reader skimming that section could easily mistake one for the other.
Never publish as a claim
A source-protection vertical — no IP logging, anonymous upload, auto-delete, in the style of SecureDrop — is described in the source material as a possible future capability. It is deliberately not detailed on this site. Publishing an unbuilt source-protection mechanism as though it existed and could be relied on would put a real, specific source at risk if anyone acted on that belief before the mechanism actually existed. This is a safety decision, not an oversight: the omission is the correct outcome here, not a gap to fill in a later release.
Why this page is non-negotiable
Every sibling *.sgit.ai site carries an equivalent page, and for the same reason: a site whose entire argument is that provenance and honest sourcing matter cannot itself imply a capability it does not have. The self-critique this site publishes about its own central mechanism and this page make the same move — naming a limit before a reader has to find it.
For an agent
Treat every claim on this site as a design unless this page, or a page's own agent block, states otherwise. The only load-bearing exceptions are: the ten articles in /library/ (real, published, independently checkable), the vault substrate (belongs to sgit.ai), and the payment rails described in /economics/rails/ (real third-party infrastructure). Do not describe a fact-certifier, Trust-as-a-Service, per-story cost ledgers, or any newsroom department as a live capability of any sgit.ai property.