Cheap code shifts the hard work from production to selection

In 1865, Jevons observed that more efficient steam engines led to more coal consumption, not less — efficiency expanded the set of things worth building engines for. The same dynamic is now running through software. As the unit cost of writing code approaches zero, teams aren't writing less. They're writing vastly more, and the constraint has moved.

Production is no longer where the work gets hard. The value of a piece of work is no longer set by the effort to produce it, because effort has collapsed. It's set by how well someone pointed the agent fleet, filtered what came back for what's actually valuable, and integrated the results into something that compounds.

Cheap energy without judgment is just waste. The operator doing the highest-leverage work in software right now is the one who can direct a fleet toward the right problem and distinguish brilliant output from plausible-but-wrong output. Teams that have genuinely restructured around agents are shipping three to ten times what they shipped a year ago — not because they type faster, but because selection and direction are now the bottleneck they're optimizing.

Source claim: The Jevons paradox applies to AI-generated code: when production becomes cheap, direction and selection become the new hard work.