The Corrected Page Now Says So

Three public records sit on this site: claims of ours that turned out wrong, rules of ours that changed, and capabilities the system gained. Four pages show you the most recent entries, read live as the page loads. Today the whole arrangement was rebuilt.

A page we corrected now says so on the page itself, with the date and a link to the correction. Until today a correction lived in our record and the corrected page carried no trace of it, so the person actually reading that page had no way to know it had ever been wrong. The notice appears only where a fact a reader could have relied on changed, never for our own machinery, and it does not expire. Two pages carry one as this is published: our methodology page, which quoted reviews nobody wrote, and one product page that named the wrong prefecture until the maker wrote to us. What it did not do was show more than one. Three days later that changed, and a page corrected twice now shows every correction, not just the newest.

Corrections and rule changes now carry a name at the foot. If the premise of this site is that one human makes every call, an unsigned correction argues with the premise. System updates like this one stay unsigned: plumbing is not a judgment call, and a signature spent everywhere is worth nothing anywhere.

Each of those four pages had been asking for the newest few entries across all three records at once. That sounds fair and is not: whichever record is busiest takes every slot. Corrections were the busiest, so three of the four pages showed three corrections and nothing else, and no rule change had been visible on any of them since 5 August. A page arguing that a system improves was proving it with three admissions of error. Each record now has a guaranteed place, newest first, and a record with nothing to say gives its place back rather than leaving a gap.

Those pages also state what the record holds now, counted as the page loads rather than typed in by anyone. A list of our own failures with no size beside it reads worse than it is.

Corrections also stopped being dressed as research. They carried a research-story label, a status line reading “Complete” and an outcome line repeating the sentence directly above it. A correction is not a study, and bookkeeping dressed as research is the small dishonesty this record exists to disprove.

Later note, 21 August 2026: the judgment this page describes, the one deciding whether a correction touched a fact you could have relied on or only our own machinery, was not being made on every correction. Nine of ours carried no judgment at all, and three pages that should have been showing a notice were not. All nine are judged now, and the count that finds a missing one runs where the correction is written.