The Jevons Paradox and the Future of Software Development
There’s an old economic observation from 1865 that keeps showing up in conversations about AI and the future of programming. William Stanley Jevons noticed that when James Watt’s more efficient steam engine made coal cheaper to use, England didn’t consume less coal — it consumed dramatically more. Efficiency didn’t reduce demand. It exploded it.
We’re living through the software version of that story right now.
The anxiety in developer communities is understandable. Large language models can write working code, debug errors, scaffold entire applications, and do it all in seconds. If the thing you spent years learning can now be partially automated, the rational fear is that your labor becomes less valuable. The pessimistic version of the story ends with mass unemployment for programmers and a hollowed-out profession.
But the Jevons paradox suggests something different might happen — and history offers a useful analogy. YouTube and TikTok didn’t destroy professional film crews. What they did was open the floodgates for vastly more filmed content than anyone had imagined possible. The category expanded so fast that there’s now more demand for skilled video production than existed before, just distributed across a much larger and stranger landscape.
Software has the same dynamic waiting to be unleashed. For most of the history of computing, the bottleneck hasn’t been imagination or even money — it’s been the sheer cost and time required to build things. Countless projects never started because the engineering hours weren’t justifiable. Countless businesses ran on spreadsheets and paper processes because custom software was an unaffordable luxury. That constraint is collapsing.
As the cost of producing software falls, demand for it won’t stay flat. It will grow to fill — and then exceed — the available supply. The contracts that never got a proper workflow tool. The small clinics that couldn’t afford custom patient management. The niche communities that wanted something built just for them. These represent an enormous latent demand that’s been waiting for economics to catch up.
This isn’t naive optimism. It’s the same pattern that played out with display technology: modern LCDs use a fraction of the power that old CRT monitors did, yet we now consume more electricity on screens than ever, because screens have colonized contexts — restaurant menus, bus stops, billboard advertising — where they were previously unthinkable.
Efficiency unlocked uses that didn’t exist before.
The more interesting question isn’t whether software development survives, but what it looks like when you’re no longer bottlenecked by the cost of writing code. Teams get smaller and faster. Individual developers own more of the stack. The interesting work shifts from implementation to architecture, judgment, and knowing what’s worth building in the first place. Those are, if anything, harder problems than syntax. There’s a genuine loss in this too. Some programmers describe a craft satisfaction in hand-writing code — the same pleasure a carpenter might feel working with hand tools. Automating that away strips something real from the work. That’s worth acknowledging, not dismissing. But the profession as a whole is more likely to look like a field that got a power tool than one that got replaced by a machine.
The vast majority of AI-generated code in the future will be written for things we don’t even do today — software projects that wouldn’t have been started, tools that wouldn’t have been justified, automations that nobody had the bandwidth to attempt. That’s what technological abundance tends to produce. Not less demand. More of everything.
Jevons saw it in coal. We’re about to see it in code.