Act I · The new production architecture
Intelligence is becoming infrastructure for assembling productive capability.
The shift is larger than task automation. Agents can discover structure, create tools, divide work, route execution and distribute capabilities once held inside institutions.
Compose the technologies before judging the industrial consequence.
Agents, open models, graph-organized work, scientific systems and physical automation are not isolated product categories. Together they begin to form a production architecture.
General reasoning can discover the structure of its cheaper replacement.
While the work is novel
General systems interpret ambiguity, explore alternatives and handle exceptions.
When structure appears
Repeated work migrates into tools, rules, workflows, state machines, caches or smaller models.
The mature demand for general intelligence may follow unresolved novelty rather than every activity it helped automate.
The economically important model is the cheapest one adequate for the outcome.
A production system can route stable, bounded, novel and critical work to different mechanisms.
The system can help produce the capabilities that improve the system.
Once intelligence reduces the cost of coding, coordination, experimentation and tool creation, it begins to attack the capital and time required to reproduce AI-enabled capability itself.
Entrants do not begin where the incumbent began.
Open implementations preserve historical work while removing some of its private scarcity.
The same capability also reaches workers, customers and small organizations. A technology powerful enough to reduce employment can reduce dependence on the institution assembling that employment.