The Hero Study Our Database No Longer Holds

In January 2026 we published a note explaining how our research citations were rebuilt around one master database instead of being typed into pages one at a time. It named a hero study for each of the four dimensions our scoring rests on. One of those studies was later removed from that database, and the post went on naming it.

We found it by instrumenting our own citations. A check now joins every citation we publish to its record in the master, and it reported this post by name. The reference was written in prose, not stored as a study reference, which is exactly why an earlier sweep, looking for stored references, went straight past it.

There were two ways to deal with it. Edit the sentence, and the record of what we published quietly becomes something else. Or leave the post standing and correct it where a reader can see the correction.

The post stays exactly as it was, and a dated correction now sits directly beneath the claim. It names the study that our database no longer holds and the studies that carry that dimension today. The study counts in the post are the counts of that January, and they stay too, because the post is a record of one day rather than a description of this one.

What this leaves behind is a rule we can apply the next time, and there will be a next time: a dated post is history and does not get quietly updated. When something inside it stops being true, the correction goes in beside it, visible, with the date it was made.