The Person in Question
William Stanley Jevons first described the paradox in 1865. He published it at the age of 29 in his influential book, The Coal Question: An Enquiry Concerning the Progress of the Nation, and the Probable Exhaustion of Our Coal-Mines. At the time, he was analyzing how James Watt’s more efficient steam engine was paradoxically causing Britain’s overall coal consumption to skyrocket rather than decline.
The Effect of the Paradox
As steam engines grew more efficient, Britain’s coal consumption didn’t fall — it exploded. Annual output rose from roughly 10 million tons in 1800 to over 100 million tons by the 1860s, a tenfold increase in just six decades. Cheaper, more efficient power didn’t curb demand; it fueled railways, factories, and steamships, unlocking uses for coal that were previously too costly to justify.
Uh? Coal? What Does This Have to Do with Software?
Replace coal with code, steam engines with AI agents — same pattern. Cheaper code should mean fewer engineers, right? Jevons says the opposite: when something gets cheaper, total demand rises, because uses once “too expensive” suddenly make sense. Cheap code won’t shrink engineering — it’ll expand what’s worth building.
If You Are a Developer, This Means…
More software gets built, not less — so more engineers are needed to build it. Your job shifts from writing every line to architecting systems, setting guardrails, and verifying AI output. Demand doesn’t vanish, it moves up the stack. The developers who thrive will be the ones directing AI, not competing with it.
If You Are a QA Automation Engineer…
You’re sitting on the era’s biggest bottleneck. As code generation gets instant, verifying it becomes the hard part — someone must catch hallucinated logic, broken edge cases, and silent failures before they ship. Engineers who can test AI-generated code at scale won’t be replaced — they’ll be the gatekeepers deciding what’s safe to release.
TL;DR
The Jevons Paradox: Just as cheaper coal didn’t cut consumption — it exploded it — cheaper code won’t cut demand for engineers either. AI makes building software near-free, so more gets built, not less. Winners shift from writing code to architecting systems and verifying AI output. Adapt now, or get left behind.