How latentwire is made

latentwire turns a news lead and the recent coverage around it into a short, attributed piece that puts the development in context — not a rewrite of one article, but the story so far.

The process

A pipeline pulls a fixed set of RSS feeds, groups items by the underlying development, retrieves the last two weeks of related coverage, and asks a language model (DeepSeek) to synthesise a brief that traces every factual claim back to a source. A person reviews each draft against those sources before it publishes. A quiet day produces nothing.

Every piece links its sources and names the subject it belongs to. Corrections are versioned at the same URL. This is AI-assisted journalism with a human editor in the loop, not automated republishing.

Contact

Corrections and takedown requests: follow the source links, or reach the site owner through the contact listed there.