August 12, 2026
The Year That Did Not Fit Our Own Sentence
Four museum works were described as all dating from the designer's years at one manufacturer. Three do. The fourth is dated four years after he left, and the clause has been removed.
One of our product pages spends a paragraph doing something we think more pages should do: it takes a claim everyone repeats about a famous object, goes to the museum’s own database, runs the search, and prints what actually comes back. Inside that paragraph, one clause was wrong.
The page said that the four works returned under the designer’s name were every one an appliance from his years at the manufacturer he trained with. The page also said, one paragraph earlier, exactly which years those were: he joined in 1955 and left in 1960. Three of the four works are dated 1957. The fourth is dated 1964.
So the sentence claimed something of all four that is true of three. The four works are real, the search is real, the dates are the museum’s own, and the correction is only to the word that gathered them.
We have removed the clause. The sentence now says the four works are every one an appliance by that manufacturer, which is true of all four, and the years stay where they were already stated correctly.
The part worth publishing is not the clause. It is why nothing caught it.
Every check we run compares one fact to the source it came from. The two facts here each passed that check, because each is correct. The error lived only in the relationship between them, and nothing we had ever looked there. A sentence that says every, or all, or since, or for twenty years, is doing quiet arithmetic on facts we already hold, and arithmetic is the one thing our checks were not reading.
The same day, on the same page, the gap turned up again in a different shape. Three sentences there joined correctly sourced facts with words that no source supplied: one asserted a cause, one described a behaviour, and one turned a single retailer into several.
That gap is now a piece of work rather than a note. Two things have to exist: a check that can find sentences doing that arithmetic and prove whether the facts underneath them agree, and a rule that makes a writer record which facts a sentence was derived from, so the check has something to read. We have written down why the rule alone would not be enough. A rule with nothing enforcing it stops being followed, and we have watched that happen here before.
The uncomfortable detail is where this was found. Not in a marketing line, and not in a page nobody reads carefully. It was in the one paragraph on that page whose entire purpose is proving we check things.