The concept is an extension of Lessig's code-is-law thesis into the domain of cognition specifically. The original thesis concerned behavior: the architecture of the internet regulated what users could do, sometimes more effectively than legislation. The cognitive constitution thesis concerns the conditions of behavior: when a tool participates in the formation of the intentions from which behavior flows, it is governing at a level prior to behavioral regulation.
The distinction between law and constitution is not merely one of hierarchy. It is one of category. Laws regulate within a framework. Constitutions establish the framework within which laws are legitimate. A law can be debated, amended, or repealed within the existing constitutional order. A constitution defines what counts as a valid argument, what rights are recognized, what powers are authoritative. When an AI tool shapes the user's sense of what connections exist, what arguments are available, and what structures are coherent, it is performing constitutional rather than legal governance.
Cognitive versus behavioral regulation. The first generation of Lessig's argument concerned behavioral regulation: the architecture of the internet determined what users could do. The second generation—the one the AI moment requires—concerns cognitive regulation: the architecture of AI tools shapes the mental framework within which intentions form, possibilities are evaluated, and decisions are made. The user who collaborates with a large language model does not merely receive outputs. The user's sense of what is possible, what connections exist, and what arguments are available shifts in response to the tool's participation. That shift is cognitive governance.
The democratic deficit of the constitutional convention. When a legislature passes a law, there is debate, public comment, judicial review, electoral accountability. When an AI company makes the architectural decisions that constitute a cognitive framework for hundreds of millions of users—what training data to use, what behaviors to optimize for, what defaults to set—there is, currently, a product meeting and a deployment. The asymmetry between the power of the decision and the deliberative process that produces it is the governance failure at the heart of the AI moment.
The smooth as constitutional enforcement. The confidence, immediacy, singularity, and polish of AI outputs are not neutral product features. They are default settings that constitute a cognitive environment: one in which uncertainty is structurally suppressed, the space for independent thought before output arrives is architecturally eliminated, alternatives are not presented, and polished output invites acceptance rather than revision. Each default is a constitutional clause, establishing the baseline from which any deviation requires the user's additional effort.