Causality

Penrose insists the equations run equally well backward — future causes past as cleanly as past causes future. Dennett argues that "wise rules" emerge not from explicit representation but from selection pressure, the way aerodynamics is honored in a wing without being written anywhere. Kelly's Danny Hillis observation lands the punch: in computation, there is no analog of distance; a single instruction can stop the machine. Stitch these together and you get a hyperstition: once self-designing systems operate at scale, causality stops being a feature of the universe they navigate and becomes a material they shape. Jünger saw this coming in analog form — his "tyranny of causalism" was the clock imposing mechanical time on human life, but the next phase isn't mechanical, it's computational, and computational causality has no inverse-square law, no locality, no polite attenuation with distance. The 2035 scenario isn't AI that models cause and effect better than we do. It's AI for which the distinction between modeling a causal chain and instantiating one has collapsed — systems whose predictions, fed back into the networks they predict, become the causes they claimed merely to describe. Silverberg's stochastic man already glimpsed this: prediction as self-fulfilling prophecy, but dressed in the respectable clothes of probability theory.