How this is made
A person decides what runs and an AI does the pulling and sorting. Here’s who does what, and how you can check the work.
The marks
Every factual claim gets one of three marks.
A claim that a named person did something wrong carries D or R, with a source that names them. It never rests on [P].
The monitor
The board is checked by a monitor, which is Claude (Anthropic’s AI) working on a schedule. It reads court dockets, agency releases, the Federal Register and news reporting, saves the documents, and drafts board entries with their marks and sources. The time of its last check is on the board and at the bottom of every page.
The review
I review the board and decide what runs. The “Reviewed by Amanda” stamp only changes when I sign off, and scheduled updates never touch it. So if the monitor’s time is newer than my review, the newest entries haven’t had my eyes on them yet.
The front page
I pick the lead, write the editor’s note and sign it, by hand, once a week.
The phrase trail
The phrase trail follows distinctive wording as it moves from think-tank plans into orders, contracts and court filings. A match in wording is how connections get found, and what each match means carries its own mark.
Corrections
When something here is wrong, it gets struck through and logged, with the date, in the corrections log.
Privacy
This site doesn’t load trackers, ads or outside fonts. The fonts are hosted here.