Research

The AI Production Function

Code Capital, Context Capital, and Computational Labor

Industrial economics models production as the interaction of physical capital and human labor, mediated by an exogenous technology term. This fits AI-native production poorly: autonomous agents perform computational work by orchestrating reusable software and accumulated knowledge rather than by selling human time — the unit of labor is a compute token, not a work-hour, and the assets that most raise output (code, context) accumulate rather than depreciate. This paper proposes a generalized production framework for AI economies — a conceptual re-grounding and a falsifiable research program, not yet an empirically validated theory. It introduces three primitives — Code Capital (C), Context Capital (X), and Computational Work (W) — derives Y = A·C^α·X^β·W^γ (which nests Cobb-Douglas), connects each to its lineage in growth theory and intangible-capital accounting, confronts the simultaneity problem for identifying the elasticities, and derives falsifiable hypotheses.