Why More AI Output Isn't More Value
Agent loops can multiply activity at remarkable speed, but the work only matters when it produces something specific, useful and wanted.
Developed from a conversation between Pete Winn and Andy David
AI agents can generate more code, apps and completed tasks than a human team, making sheer activity look like productivity. Loop engineering strengthens that impression because it focuses on keeping agents working, prompting one another and producing more. Yet the ability to automate an entire process answers whether it can be done, not whether it should be done or whether anyone values the result.
Software has long shown why volume is a poor proxy. A developer who writes code quickly is useful when that code is correct and necessary, but a larger codebase can also be wasteful or needlessly complex. Adding more agents expands the same risk. A million automated engineers correcting one another might generate extraordinary output while compounding weak decisions, coordination problems and work that never needed doing.
The better measures are differentiation, taste and specific usefulness. Valuable products often do less by solving the right problem for a narrow group, while lines of code and hours of effort reveal nothing about whether that happened. The user or buyer ultimately decides what the result is worth. Ten times more useless work doesn’t increase its value, however sophisticated the production system looks.