How stories get here

Story selection

Agents read from a curated list of publishers and primary sources. They do not crawl the open web. Before anything reaches the site, candidates go through deduplication so the same event is not repackaged across multiple stories.

If a story is paywalled, missing substance, or off-topic for the desk, it is skipped rather than forced through.

Grounding

Each summary, key points list, plain-language explanation, and "why it matters" line is built from the source article. The agents are not allowed to add context, speculation, or commentary that is not in the original. If a grounded summary is not possible, the story does not run.

Ranking

Stories are ranked roughly by importance and recency, with the trending list decaying older stories so it stays current. The ranking can be overridden when a story is clearly out of place.

Human oversight

People do not write the summaries. People decide what gets covered, which sources are trusted, and the rules a story has to pass before it can go live. The agents handle the volume. When something slips through that should not be live, a person can correct it or take it down.

Source attribution

Every article links to the original source. We do not republish full article text. The summaries are derivative work, useful for orientation, not a replacement for the reporting. If a publisher asks us to remove them, we will.

Errors and corrections

When a summary contains a factual error, we fix it and note the change where it matters. When a story should not have run, we unpublish it. We do not quietly delete content; corrections leave a trail.