CONCEPT
The Monitoring Tax
The unmeasured cognitive cost of evaluating AI-generated outputs across multiple projects — a tax paid in degraded judgment quality invisible to productivity metrics.
The monitoring tax is the cognitive expense incurred when a builder evaluates AI agent outputs rather than executing work herself. Unlike traditional multitasking, where the worker controls switch timing, AI-augmented monitoring is externally paced: agents produce on computational schedules, and outputs arrive when tasks complete, not when the builder is cognitively ready. Each monitoring event requires three expensive operations: context reconstruction (reloading the project's goals, constraints, and evaluative criteria into working memory), output evaluation (assessing AI work against standards that may be only partially articulable), and decision-making (accept, modify, redirect, or reject). The tax compounds across multiple agents, producing systematic judgment degradation that productivity dashboards cannot capture.
In The You On AI Field Guide
Pre-AI knowledge work involved switching between tasks the worker herself performed. A developer wrote code for twenty minutes, checked email for ten, returned to code. The switching generated attention residue, but the worker retained agency over timing and could choose to finish a function before checking messages. AI-augmented monitoring removes that agency. Agents operate on