AI can accelerate a newsroom. It cannot be the newsroom.
This site drafts with machine assistance and publishes nothing a human editor has not verified against primary sources. Here is the workflow, the policy it answers to, and the line it will not cross.
The short answer
Briefings on this site are drafted with AI research assistance and published only after a human editor verifies every figure against the linked primary source. Each briefing carries a production note saying so. The desk treats machine assistance the way professional kitchens treat sharp knives: standard equipment, dangerous exactly in proportion to the carelessness of the hands.
Why disclose at all
Two reasons, one ethical and one practical. Ethically: a site whose entire argument is “follow the sources, not the slogans” forfeits its licence the moment its own production is opaque. Readers deserve to know how the sausage is drafted, which is why the About page names the editor, the point of view and the conflict of interest in the same breath. Practically: search engines now police this frontier explicitly. Google’s spam policies define scaled content abuse — generating many pages primarily to manipulate rankings rather than help users — as a violation regardless of whether humans or machines produced the text. Its March 2024 update, which introduced the policy, was aimed at exactly the flood this site refuses to join.
The workflow, step by step
Research: the editor sets the question; machine assistance gathers candidate sources and data. Verification: every statistic, quotation and claim is checked by the editor against the original document — the survey PDF, the policy page, the court record — never against a summary of it. Sources that cannot be verified are cut, and the briefing says less as a result. Drafting: assisted, under the site’s style and evidence rules. Editing: human, line by line, with the source ledger assembled alongside. Publication: with a production note and a dated corrections route. What the machine may not do is establish an uncited conclusion. If no source survives verification, no claim survives editing.
The line this creates
Volume is the tell. This site could publish fifty briefings a day; it publishes few, because verification is the bottleneck and the bottleneck is the product. An operation that removes the bottleneck — auto-publishing summaries of summaries at scale — is running the business model Google’s policy describes and this desk’s methodology forbids. The difference is not the tool. It is whether a human stakes their name on every sentence before it ships.
FAQ
Why not publish at machine speed, since the tools allow it? Because this site’s value is the verification layer, and verification does not scale with token counts. Fifty unchecked briefings a day would make the desk the very thing it audits.
What happens when the machine is wrong? The same thing that happens when the editor is wrong: a dated correction, in public, on the page. The production note exists so that responsibility has an address.
Will readers ever see fully automated content here? Automation may assemble, format and translate. Conclusions, verdicts and every number in a source ledger require a human signature first — structurally: the CMS will not publish a briefing without its production note and source list.
The bottom line
Machine-assisted, human-verified, publicly correctable — and always disclosed. Judge the briefings by their sources, and when a source fails, use the corrections desk; the record changes in public.
Written and edited by the desk, then verified with AI assistance: every figure, quotation and link in the source ledger was checked against the original document before publication.