The Number We Stopped Typing By Hand

Until today, a reader on one of our product pages saw one account of how many data points we check, and a machine reading our published files saw a different one. The machine files had been counting the total from the field definitions this site actually registers, ever since our last pass over that layer. The pages carried a figure typed into the template by hand.

Both were meant to describe the same thing. Only one of them could ever notice when it stopped being true. A typed number does not know the site changed around it.

So the product pages, the badge on every lab note that passed, and the research descriptions those pages hand to search engines now read from the same counter the machine files use. The total is measured when the page loads, from the field definitions a product record carries, counted once and held for half a day before it is counted again.

The counter is allowed to come back with nothing. If it cannot reach those definitions during a deploy, it returns an empty result instead of a guess, and every sentence that would have carried the total falls back to wording that carries none. We would rather say less than publish a figure we did not count.

One line changed shape entirely. A lab note title used to name the total. Titles are the most copied and least maintained sentences on any site, so a count no longer appears in a title here at all. It belongs in the body, where it is counted, or it does not appear.

This did not fix every page. Several of our explainer pages still describe the system with a hand typed figure, and each of those is queued for a pass of its own. We are publishing this now rather than when that sweep finishes, because the gap between what our files said and what our pages said is the kind of thing a reader should hear from us first.