Note
Everyone predicts in January and finds out in December
The clustering is severe enough to be a structural fact about the genre, not a quirk of one archive.
Twenty-nine of the predictions on this record settle on the same day: 31 December 2026. Not near it. On it.
They come from sixteen different people who have never coordinated with each other — a UX researcher, an election forecaster, two venture investors, a technology columnist, several AI critics. The clustering is not a quirk of what got collected. It is a fact about the genre.
Why the calendar looks like this
Public forecasting is an annual ritual. People publish predictions in December and January, because that is when readers want them, and they write “by the end of the year” because that is the horizon a year-opening piece naturally takes. The deadline is inherited from the publishing convention, not chosen to fit the claim.
Which means the resolution calendar is lumpy in a specific way: almost nothing settles between February and October, and then everything settles at once. Between now and December this record has exactly three results arriving — one in November, two before that. Then twenty-nine in a day.
What that does to accountability
It makes it seasonal, and seasonal accountability is weak accountability. A prediction made in January has eleven months of being interesting followed by one day of being judged, and by then the conversation that produced it has moved on entirely. The people who were loudest in January are rarely in the room in December.
The fix is not to nag people into shorter horizons. It is to make the December date exist as a thing you can point at in advance — a fixed appointment, published, with everyone’s name attached, counting down in public. The claims were made in the open. The reckoning should be scheduled in the open too.
That page exists and updates itself. Right now it reads as a list of what sixteen people believe. On 1 January it becomes a scoreboard, without anybody editing it, because each test was fixed before the answer was known.
If you know a dated call that belongs alongside them, put it on the record before the answer is in. Afterwards it still gets kept, but it is filed as historical and never counted in anyone’s track record.
Every figure above is a count from the record on August 29, 2026, and the archive is exportable if you want to check the arithmetic yourself.
Also here: People who write predictions get scored. People who say them don't. · What makes a prediction testable · The goalposts problem · What a 96% accuracy score actually measures · How a track record lies without a single false statement · How to check whether someone actually predicted what they say they did · Sometimes the honest verdict is that nobody can tell