I read "The Lessons of History" these two days. It moved me, and then it did not convince me. It is an intelligent book, but its intelligence is empirical: it takes what has held for a few centuries, generalizes it, and hands the result to the future as law.

The problem is that induction has exactly one precondition — that the future's set of possibilities stays inside the past's set of samples. In a world where the density of technological change is accelerating, that precondition is precisely what fails. The changes of a single century have already left the imagination of everyone who came before: not the same events happening faster, but types of events with no ancestor in the sample at all. To predict the consequences of AI or controlled fusion using a library that has never seen either is to produce error with very high precision.

I am not dismissing history. It tells me how people behave under pressure, and that mechanism is stable. But it does not tell me what shape the world will take. The first is human nature; the second is structure. Samples of human nature can be reused. Samples of structure cannot.

What interests me more is the structural direction of this era. Technology has raised the cost of controlling other people and lowered the cost of growing alone. When anyone can run an agent on their own machine that serves only them, when compute and models no longer have to sit in a few server rooms, humanity fades out of the field of view of those who want to control everything — not saved, just no longer worth controlling, and no longer controllable.

So I will not price the future with history. History can give me courage and alertness. It cannot give me a map.