August 20, 2026
The Citation Read a Number Out of Our Own Sentence
The citation box on lab notes and where it was getting the score it published
Every lab note carries a citation box. It is there so a writer or a research tool can quote what we found and attribute it properly, and on a note about a product we rejected it was built to include the score.
We never stored that score anywhere. So the citation went looking for it in the only place a number existed: the one-line summary at the top of the note, written as prose by a person. It matched the first thing in that sentence that looked like a score and offered it as ours.
Across the whole library that pattern found a number exactly once. The note was Fifty Stockists and Zero Reviews, and its summary line ends with a confidence reading of seventy out of a hundred. Confidence is not a score. It is our own measure of how much evidence we managed to gather, and it runs on a different scale entirely. Anyone who copied that citation carried away a number that had never been a verdict on anything.
The same pattern was silent on all one hundred and ninety eight rejection notes, and on eight of them it was silent while the note itself stated a figure a reader could see, because those sentences put the word before the number and the pattern expected it after.
What changed
We deleted the pattern rather than improving it. A better version would still have been guessing, because a rejection note discusses four different scores and a pattern cannot tell which one it has found. A wrong number inside a citation does not stay on our page. It travels into somebody else’s document under our name.
A rejection now has a field of its own that holds the one figure that matters, the weighted total the product actually reached. The citation reads that field and nothing else. Until we have written those values, one for each note and each one traceable to what that note already published, the citation carries no score at all.
Why the empty is the honest state
Between now and the day those values are written, a rejection citation will show the verdict without the number. That is a real loss and we are choosing it. The alternative is a figure recovered from a sentence, which looks exactly like a figure read from a record and is right only by luck.
The same version retires a row from the filing card at the foot of each note. It repeated the headline directly above it, and on a passed note it repeated the product card as well.