AI is exposing that the market, money, the firm, and the job were compressions: coordination forms shaped by what was too expensive to compute. That cost just collapsed.
The argument is information-theoretic, not ideological. Rich distributed coordination (many parties, many value dimensions, continuous mutual evaluation) was always conceivable and always too informationally expensive to run. So economies compressed. The price signal reduced value to one public number; money reduced obligation to one fungible unit; the firm absorbed the coordination overhead a market could not price; the job packaged human capacity into a unit the firm could account for. Each compression was rational under its era's computational constraint.
The LLM-driven agentic stack collapses that constraint. When the cost of computing rich coordination falls toward zero, the compressions lose their justification, and economic protocols drift toward whatever distributes the most economic agency to the edge. This is a bet on capacity, not ideology: no one has to be persuaded of a program, any more than firms were persuaded to adopt spreadsheets.
The public anxiety is real but mislocated. The question "will AI take the jobs" assumes the job survives as the unit of economic participation; the deeper event is that the coordination layer the job belongs to is becoming legacy. The productive question is the one this project owns: the economy has always been a computation; AI just made it programmable. Who gets write access?
The register discipline: this critique is economic, never moral. It does not claim AI is intelligent, does not moralize against markets or firms, and does not prophesy collapse. It observes a constraint expiring and asks who authors what replaces it.
Related: · · ·