The four engines
Z-M Intelligence runs four engines that feed one shared library and one shared paywalled portal.
1. APAC Policy Monitor
The Monitor watches authoritative government sources across Asia-Pacific. Each source is described in a YAML registry — its canonical URL, the index where new policies appear, the cadence the source updates at, the rate limit we respect, and the licence position recorded at onboarding. A scheduled runner sweeps each source on its declared cadence (hourly, daily, weekly), parses the new or changed policies, computes a SHA-256 hash of the canonical text, and writes a new version row whenever the hash differs.
Members see the current version, the full version history, and a plain-English summary of what changed between any two versions. Every version has a citation-stable permalink that never breaks.
2. Source Digest
The Digest tool turns a YouTube video, a PDF, or an HTML article into a normalised, neutral, thesis-style summary — never "in a recent video", never "at minute twelve". The medium is invisible to the reader; the substance is what travels. The output feeds the Director Voice Writer and, when the source is independently valuable, can be published as a standalone Source Brief.
3. Director Voice Writer
The Voice Writer composes one or more digests plus an editorial brief from the director into a long-form article in his established voice. It is not a generic AI rewriter — it is bound by a written voice guide derived from the director's existing writing, with hard lint checks for banned terms and forbidden phrasings. The director reviews and approves every draft before publication.
4. Intel Scout
The Scout is the discovery half of the Monitor. Where the Monitor watches known sources, the Scout searches government RSS feeds, university repositories, think-tank publication indexes, and curated news feeds each day for new candidate documents that match the library's pillars. Candidates are scored, deduplicated, and queued for the director to triage in the admin console.
The rights pipeline
Every document and every policy version carries a documented rights stance. The library uses three tiers:
- Tier A (host). Public domain, CC0, or clearly-public-record material from US federal works, pre-1931 US works, and explicitly-PD library scans. We host the file and label it accordingly.
- Tier B (host with attribution). CC BY 4.0 and Open Government Licence material from data.gov.au, NZGOAL, World Bank Open Knowledge Repository, UK GDS, Canadian Digital Service, and similar. We host with prominent attribution.
- Tier C (link-only). Material we do not have a clear right to host. We index it (title, abstract, outbound link, rights note) but never copy the file. The detail page links straight back to the rights-holder.
Where rights are unclear, the default is Tier C. Better to link than to host material we cannot defend. The full intake pipeline — discover, verify, classify, ingest or index, editor's note, publish — is documented in our internal SOP.
Defence content — the line we draw
Defence-adjacent material is the easiest place for an intelligence library to lose its credibility. We accept material that has been formally cleared for public release — DSTG technical reports, ASD/ACSC public guidance, NIST and CISA joint guidance, parliamentary committee reports, published Defence whitepapers, and similar.
We refuse material that is restricted, partnership-only, accessed via member networks, classified at OFFICIAL: Sensitive or higher, FOI'd without subsequent public-release posting, or circulated as AUKUS trilateral working papers. We refuse it regardless of how a copy was obtained, including by people who would prefer we publish it.
Director sign-off is required before any Defence-adjacent entry is published. This is non-negotiable and applies equally to material the director himself encounters.
Sovereignty
The Library is operated under Australian jurisdiction. The authoring pipeline runs on infrastructure under our control with Australian-residency assurances on the underlying LLM providers. Phase 2 of the build moves the entire stack to fully sovereign Australian cloud — Vault Cloud or AC3 — at which point all customer data, document text, and authoring traffic operates exclusively within Australian jurisdiction without foreign-cloud dependency.
We do not access adversary-jurisdiction government infrastructure from Z-M IPs. Coverage of those jurisdictions comes via trusted third-party analysis under Tier C, never via direct scraping.
Verifying anything we publish
Every entry in the Library carries a link to its original source. Every policy version carries a link to the government's authoritative copy. Every Insights article carries its References block at the end. If we are wrong, the mechanism for proving we are wrong is built into the page.
See the source registry for the full list of repositories and institutions we draw from, with rights position recorded for each.