One of the 47 came true. Here is what happened, and what we do with the number we had.
One of the 47 theses on the map has come true. It is the first.
The thesis said an electricity bottleneck would slow the build-out of AI data centres before 2028, and its criterion asked for one of three things: an announced cancellation for lack of grid access, a regional moratorium, or a connection auction with a wait of more than five years. On 14 July 2026 the second one happened: the Governor of New York signed an executive order halting environmental permits for a year for hyperscale data centres seeking to connect to the state grid.
Two other things happened the same day. Neither resolves the thesis on its own, but both point the same way: CAISO — California's grid operator — recorded the first of three consecutive days of extraordinary demand, and Texas tightened its own moratorium.
What we do with the number is the part that deserves explaining. The day before the executive order, Brújula had this thesis at 93%. That 93% stays. We are not raising it to 100% now that we know the answer: we label it last forecast and leave it where it was.
The reason is simple. The only interesting thing about a resolved thesis is how confident we were before we knew. If we rewrote that number, the Calibration Yearbook — where we publish each year whether we got things right — would be measuring our forecasts against figures we wrote once we already had the answer. That is not calibration; it is marking your own homework.
The thesis stays on the map, flagged as resolved, with the date, the outcome and a link to the source. Its master thesis — the physical-infrastructure bottleneck — rises by a point, because it now folds in the result instead of averaging a doubt that no longer exists.
One note on the criterion. The clause about electricity demand said «sustained» without saying for how long. We defined it — three consecutive days above the threshold in the same region — before resolving, not after. A heatwave touching five regions for one day each is not a grid problem.
