What a machine-written wire can and cannot do

The Dorworth News Service writes its stories by machine. A model reads a primary source, such as a filing, a ruling or a data release, and drafts a short story about what it says. That makes the wire fast and tireless, and it makes the limits of the method worth stating plainly.

What the wire can do

The wire can read a long document, report its contents and cite the document it drew from, then do the same for the next document without tiring. Because it draws only on primary sources and never on other news outlets, every claim in a wire story can be traced to a document the reader can open.

Before release, every story is checked by a panel of AI models, each running a separate test: whether the story is faithful to the source, whether every number matches, whether the tone is neutral and whether the story is consistent with the record. Every story goes through every check; the panel does not sample. A story that fails any check is held and is not published.¹

What it cannot do

The wire cannot judge what a document leaves out, and it cannot tell a reader whether a source is right. That is why every wire story carries a bias note: a statement of what the source can and cannot show, so a reader knows the limits of what they are reading.

And the wire does not answer for its mistakes; people do. Corrections are decided by the desk, made on the story with a note explaining what changed and listed at /corrections. Readers who see an error can write to desk@dorworth.com.