Editorial
Editorial standards
Briev is written by software, so its standards are engineering: what gets read, how a story is written, and the checks that run before anything is published.
What Briev is
Briev is an automated newsroom. Software reads the news market whole — every side, every hour — and writes one clear account of each story, plus an analysis of how that story is being covered. These are the standards that pipeline is engineered to meet. They are enforced by design, not by promise: where a rule can be checked by machine, it is.
Reading the whole market
Briev continuously reads hundreds of newsrooms chosen to span the political spectrum — left, centre and right in deliberate balance — plus wires, trade press and science desks: tens of thousands of articles, around the clock. The source list is curated for breadth and factual track record, and rebalanced as outlets come and go. No single outlet, however large, determines what Briev treats as true or important.
One story, many accounts
Every article is compared against recent reporting and grouped with the other write-ups of the same event. A story on Briev is therefore never one outlet's account: it is the event as reported by everyone covering it, and the coverage counts you see reflect that directly.
Writing, and re-writing
Briev's text is synthesized from the full group of reports — never copied, never lightly paraphrased from any single article. As a story develops, its account is re-written from what the coverage now says, and a development log records how it moved. A story's page should describe the story, not the hour it first broke.
Accuracy guardrails
Verification before publication. Every generated account is checked, claim by claim, against the underlying reporting before it replaces anything. A second, independent automated checker reviews the same text — one checker is never trusted to grade its own work.
When in doubt, the old text stands. A rewrite that contains a claim the reporting does not support is repaired; if it cannot be repaired, it is discarded and the last verified version keeps its place. Publishing nothing new beats publishing something wrong.
Numbers are special. Casualty figures, counts and dates are held to strict checks: they must be traceable to the reporting, disagreements between outlets are tracked rather than averaged away, and unsupported figures block publication.
Ranking and alerts
Stories are scored for significance against a fixed editorial rubric, then re-calibrated against the rest of the day's news, so a slow Tuesday cannot inflate a routine story into a headline. Breaking-news alerts have a harder gate: confirmation from multiple outlets on more than one side of the spectrum. A single newsroom's say-so is never enough to push a notification.
Coverage analysis
Political lean is assessed at the outlet level, using coarse public third-party ratings — never per article, and never by us deciding what an article "really" means. Flags like blindspot and underreported describe patterns of coverage: who is running a story and who is not. They are statements about attention, never verdicts about truth.
What never influences selection
Briev carries no advertising and sells no placement. Nothing about what is covered, how it is ranked, or when an alert fires is influenced by engagement metrics, commercial interest or any party's preference. The scoring rubric is the same for every story about every subject.
Limits and corrections
This is automated analysis and it can be wrong: a mangled figure, a story grouped with the wrong event, a framing summary that misses the point. We publish that caveat on the site itself, and we mean it. When you find an error, tell us at [email protected] — fixes are made to the pipeline, not just the page, so each correction makes the next thousand stories better.