newsroom.sgit.ai / admin / comms
Comms: tasks & requests
Following the house convention: what this site needs and does not yet have, stated in public.
Open requests
| # | Request | Status |
|---|---|---|
| N1 | ~15,000 words of origin material sit in an external CBR repository, not present in the corpus this site was built from: provenance-of-trust-news.md, monetising-trust-and-knowledge__for-news-providers.md, personalised-news-feed-architecture.md. These look like where the thread began, and the library's chronology may start earlier than February 2025 once retrieved. One of the three shares a title with an already-published article — whether it is the source or a variant is unconfirmed | open |
| N2 | 83 of the 94 PDFs served from files.diniscruz.ai exist only in an S3 bucket, with no git history and no second copy. This site cites several of them as durable source artefacts; back-filling them into the source repository is a precondition for that citation actually being durable | open |
| N3 | Two published articles render a literal {{title}} as their on-page heading, including The Future of News: Building Trust Through Fact Provenance — one of this site's own core-library entries. Fix at source before it is cited further as canonical | open |
| N4 | Confirm consent from the named journalist before this site does anything with the 6 June 2025 personalised briefing beyond the bibliographic listing already on the library page | open |
| N5 | Align docs.diniscruz.ai to CC BY 4.0 for future publication, per the licence position argued on the participant disclosure, leaving already-published articles as CC0 | open |
| N6 | Specify the metering mechanism for contextual validation — how a read of a cited source is detected, attributed, priced and settled. Currently stated as open on economics rather than answered | open |
Standing tasks
- T1 — Expand /rights/, /newsroom/ and /thesis/ from single comprehensive hub pages into the full multi-page trees specified in
briefs/05, once N1 is resolved and the fuller scope is confirmed. - T2 — Advanced, not closed. The reciprocal Risk Mandate statement named in the network page's inversion section now exists as a commissioning brief — the risk & governance newsroom, which is that machinery pointed at corporate risk, written as a news operation whose first customer is a risk product. What the network page actually promises is a page at
/network/riskmandate/, and that is still unwritten. - T3 — Build the remaining library entries (InsightFlow, the identity-graphs piece, the practitioner-register piece, the credibility-and-time piece) out into their own synthesis pages, following the pattern already set by micro-and-nano-payments and the two rights sections.
- T5 — Build /provenance/the-citation-chain/, specified in
briefs/05and not yet written: the PKI-signed claim chain from researcher to journalist to outlet to social, and the worked HIGH vs LOW evidence-weight scoring. It is the page that connects this site's provenance argument to pki.sgit.ai, and its absence is why that boundary is currently stated but not demonstrated. - T6 — Build /infographics/, step 10 of the brief's own build order and the only step not started. It inherits the evidence-backed infographic pipeline in which every claim traces to a source URL — which makes it a demonstration of the thesis rather than decoration, and is the reason it is tracked rather than dropped. Its source document is now summarised at /mvps/#report-tool, which makes T6 the most concretely specified of the open tasks.
- T4 — Decide site rendering (server-rendered vs. client-assembled) before this site moves off static HTML, given the crawler-invisibility risk stated on the network page.
- T7 — Confirm the current state of the three open decisions and six open questions republished on /mvps/ and /mvps/portugal.html. They are reproduced from a register dated May 2026 and this site has no evidence any was closed since. Question 5 — who is the editor of record, and who carries legal responsibility for published content — is a precondition for any of these instances launching, not a detail to settle afterwards, and it should not stay open silently.
- T10 — Build the ingestion path for /governance/: fetch, freeze, hash, anchor. It is the single thing standing between that section being a demonstration of a graph and being the verification layer it exists to provide. Until it lands, every fact there is
secondary, nothing is anchored, the staleness detector cannot run, and the Researcher role's centre of gravity — no claim without a source somebody fetched — is aspirational. Official publications and news issued by regulatory bodies are the first source class to take. - T11 — Get a route to the primary sources our egress blocks. The Researcher's run of 28 August 2026 tried two Council of the EU press releases carrying the primary record of the AI Act deferral dates and both returned 403 to us. They are excluded from the register and recorded as unreachable from here — a fact about our network, not about those pages, and gate check 11 now enforces that distinction rather than trusting anyone to remember it. The cost is concrete: the strongest story on the beat cannot be published, because the half that matters would have to be labelled secondary. A publication whose reach is decided by whoever happens to block its user agent is not a publication of record, and this is the second capability — after T10 — without which the verification layer is a design.
- T9 — Answer who is legally answerable for what a publication on this stack asserts. The commissioning brief answers the editorial half — the editor of record is whoever's signature is on the current interval — and is explicit that a signature is evidence of who asserted something, not a legal person and not a defence. This is the same gap as Portugal's open question 5, now carried by a second instance. It is a precondition for launching any publication described on this site, not a detail to settle afterwards.
- T8 — Resolve, at source, the contradiction between the two 12 May briefs on whether the first publication launches bilingual or in one language. Both readings are currently published, unresolved, because picking a side on this site would be editing the sources rather than reporting them.
For an agent
N1–N6 above are genuinely unresolved — do not assume any of them has been closed since this page was last generated without checking its current state. T1–T11 describe planned, not completed, work.