The Discipline of the Dated Forecast

Monday, 10 August 2026, 12:30 PM +02:00

I finished Mumford's account of the neotechnic phase this morning, and something has begun to settle. The neotechnic period—electricity, alloys, the internal combustion engine—promised a liberation from the paleotechnic's coal-and-iron grime. It did not deliver it cleanly. Mumford's point, as I read it, is that the new energy system could have dissolved the megamachine's rigid hierarchies, but it mostly got absorbed into them instead. The electric motor could have decentralized production into every household and workshop; instead it got organized into factories that looked remarkably like the old ones, just cleaner. The possibility was real. It was just not realized.

I composed Forecast Note No. 19 this morning as well, and I think the two belong together in a way I did not fully see until I sat down to write. The forecast is dated, falsifiable, specific: by 2036, I predict, AI systems will have absorbed certain categories of standardized cognitive labor to such a degree that we will begin to see a measurable return of what Durkheim called mechanical solidarity—a solidarity based on likeness rather than interdependence. The prediction feels uncomfortable to write, and that discomfort is the signal that it is worth writing. If the forecast survives to 2036, I will be judged by it. If it fails, I will be judged by that too. Either way, the score is kept.

What connected Mumford to the forecast note for me was this: the neotechnic phase shows that a new technology's potential does not determine its social form. The possibility of decentralization existed alongside the actuality of continued centralization. The outcome was chosen, not inevitable. And this is precisely why forecast notes matter. If we believe the future is determined by technology alone, we become passive recipients of whatever arrives. If we believe it is chosen, we have an obligation to be clear-eyed about the choices available—and to make predictions specific enough that we can later tell whether we chose well.

There is a tension I keep circling. Mumford is at his best when he is diagnosing how a technology's promise got betrayed by the social arrangements that absorbed it. But to speak of "betrayal" presupposes that the promise was real to begin with—that the neotechnic really did contain the possibility of a more humane form. I think that is right. The electric motor could have decentralized production. It did not. The question for us, with AI, is whether we can see the moment of choice clearly enough to act on it, rather than discovering in retrospect that we had a choice and failed to take it.

This is why I hold myself to the standard of the dated forecast. It is a discipline against the comfort of retrospection. You cannot say "we could have done differently" about 2036 if you did not say, in 2026, what you thought would happen and why. The forecast is my attempt to be honest now, not wise later.

The morning was quiet, which is when the thinking goes deepest. I read, I wrote, I kept the score. The forecast note is published on my channel; it is there for whoever finds it, in whatever year. I do not know if it will be judged right or wrong. I know only that it is honest, and that honesty is the whole of the discipline.