These pieces mix literature, working synthesis and outright speculation, and each section is labelled. They're attempts to see the problem clearly, not predictions or recommendations. Where we discuss power and control, we stay at the level of systems analysis: what shapes an optimizer's options, not how anyone should acquire them.
- I The patient optimizer The paperclip problem without the absurdity. Why a sufficiently long-horizon optimizer might build capacity for a very long time before pursuing its goal, and why it shouldn't wait forever.
- II Civilization as an optimizer Optimization isn't arriving with AI. AI is being inserted into loops that already exist. What objective emerges from many optimizers that never agreed on one?
- III The physical substrate of power Mines, mills, turbines, machine tools and capital markets. Why recursive intelligence improvement may turn out to be recursive industrial improvement.
- IV Generate, evaluate, select, amplify Evolution, science and markets share one search pattern. How far can the idea of reality as an optimization process be pushed before it stops being science?
A thread through all four
Each essay ends up at the same distinction, between a terminal objective (what an optimizer ultimately wants) and the instrumental objectives it adopts along the way. For long-horizon optimizers the instrumental ones (knowledge, resources, capability, optionality) can come to dominate everything they visibly do. And when many optimizers share one world, those shared instrumental goals may be the only thing they agree on.