This piece traces a simple but unsettling economic lesson: when tools make something cheaper to do, human demand usually expands rather than contracts. It follows the arc from a 19th century observation about coal through computing’s history to the present AI surge, showing how efficiency often fuels more activity. The aim is to underline why this matters as AI lowers the cost of trying and expands the scope of what people decide to build.
In 1865 William Stanley Jevons noticed a paradox: better steam engines did not shrink coal use, they multiplied it. As engines became more efficient, coal grew cheaper to burn and industries simply used far more of it. That pattern later became known as the Jevons Effect and it still rings true in new contexts.
The core idea is blunt and practical: whenever a resource or process gets cheaper and demand responds to price, people find new ways to consume it. Efficiency does not automatically translate into leisure or conservation. Instead, lowered costs open opportunities and those opportunities tend to be filled.
We face a comparable inflection with artificial intelligence, a set of tools that promise to speed up knowledge work the way machines sped up physical labor. The sales pitch says AI will free us from drudgery, automate basic tasks, and gift us more free time. Reality, if history is any guide, will probably be messier and more work-filled.
Back in 1930 John Maynard Keynes predicted a short workweek by the early 21st century, imagining abundance would let people choose leisure. That did not happen; instead productivity translated into more goods and services, not fewer hours worked. We repeatedly took the gains as increased output rather than shorter toil.
Computing offers a clear modern prologue. Mainframes were once scarce and expensive, reserved for massive firms, but falling costs and better tools spread computing everywhere. Each wave of cheaper, more capable tech—from minicomputers to personal machines to the cloud—made computing more ubiquitous rather than obsolete.
When programming moved from low-level tedium to higher-level languages, developers did not disappear; they wrote far more code and tackled bigger problems. Open-source libraries and cloud platforms made development easier and cheaper, but easier tools expanded the set of things people attempted. The effect was an explosion of software across industries rather than a contraction of developer labor.
Now LLMs and coding agents cut the “cost of trying” even further, letting tiny teams or solo founders attempt projects that once required many people. Work that needed a whole squad can be initiated by a single person experimenting with agent workflows. Lower barriers lead to more experiments, more projects, and more ongoing work to manage those projects.
Consider the case of an engineer who used an AI coding assistant to submit hundreds of pull requests and change tens of thousands of lines in a month, all driven by that assistant. That story is not about jobs disappearing into silence; it’s about one person scaling output to match a group. “The result is not a workforce at rest.” The capacity to do more stretches into more things being done.
As tasks shift, human roles tend to evolve toward orchestration and oversight rather than simple production. People become managers of fleets of agents, curators of outputs, and designers of goals for automated systems. The span of control for individuals expands, but the work itself proliferates as newly possible projects get green-lit.
In cultures that prize growth, like the United States, efficiencies routinely turn niche tasks into mainstream expectations, creating whole new job categories. Marketing is a perfect example: tools made many parts of the job easier, yet marketing employment grew dramatically as those tools enabled businesses of all sizes to compete in content, data, and advertising. Easier tools broaden demand rather than narrow it.
The real danger lies in confusing ability with wisdom. Just because a task can be done faster or cheaper does not mean it should be done at all. As humans move into judgment roles—deciding which projects agents should pursue—we must insist on priorities that favor meaningful outcomes over busywork and noise.
Technology will keep pushing down costs and opening frontiers. That dynamic hands us power to build more, but it also hands us responsibility to choose what to build and why. Efficiency will not decide those questions for us.
