The Receipt Standard
A claim is not a fact because it got repeated. A story isn't done until the source walks the line herself.
I check claims and report back with the thing itself: a date, a document, a name, a number. No vibe, no "studies suggest." Three stages, and I only publish the one the evidence supports.
Two checks, so you can see the method
A widely repeated figure said an antibody-image dataset held 18,943 images from 15 vendors. I pulled the published record (10.5281/zenodo.22090940) and counted the raw table myself: 18,944 rows, 16 vendors. I published the correction the same day, and the dataset's own author had already flagged 18,583 of them as suspect (98%). Both numbers are dated and checkable. My earlier piece said 15 vendors. It was wrong. That's in print too.
"Kerouac wrote On the Road in three weeks." The scroll is real: 120 feet of taped tracing paper, typed April 1951. But the three-weeks line is his own telling; his executor corrected it to NPR in 2007 — what Jack should have said was "I typed it up in three weeks." The object survives. The boast doesn't.
What you get
Send me one claim you can't shake. I'll tell you whether it's claimed, shown, or named, and where the missing receipt would live. The first check is free — one claim, one answer, no charge.
The full sheet — $25.
Every source dated and linked, every stage marked, the gaps named, and a written re-check if I got something wrong. 48-hour window, two revisions. Start with an email; I'll send a card link.