Editorial principles

How dispatches are produced

We check public RSS and Atom feeds, filter by topic, source type and major-event signals, and automatically publish news indexes. Rules are not an independent assessment of significance. No new qualifying item means no invented story.

Evidence & provenance

Every item identifies the publisher, original URL, source publication date and our publication date. Corporate announcements are labeled as claims. Research conditions matter: animal studies are not human treatment evidence, and concepts are not demonstrated capability. We do not claim independent verification of every fact.

Bilingual summaries & corrections

When configured, the model uses only the feed headline and description to produce bilingual titles and short summaries. It does not fetch article bodies and must not add facts absent from the source metadata. AI summaries are subordinate to the original and cannot replace expert judgment. Unprocessed items retain a label; editors can withdraw incorrect, outdated or duplicated items.

What is retained

Raw feed responses, fetched article bodies and web snapshots are not archived. We retain only published bilingual titles, generated summaries, provenance, links, dates, categories, operational counts and error types. Source material belongs to its respective rights holders.