When AI coding assistants started taking over parts of software development, the panic was predictable. Programming was dying, engineering teams would shrink to nothing, writing code was about to be obsolete. Here in 2026, something very different is happening.
There is an old idea in economics called the Jevons Paradox. When a resource becomes more efficient to produce, consumption does not fall. It explodes. Cheaper coal in the 1800s did not mean less coal burned. It meant coal got used for far more things, so total demand went up. The same thing is happening to software. As the cost of building drops, the amount of software we build rises to fill the space, and then some.
I did not read about this. I lived it.
From zero to more than I planned
For sixteen years I worked in IT without launching a personal site. A custom website always felt like a project that needed dozens of hours I could never quite justify against everything else. The math never worked, so it never got built.
Then I used AI agents for the initial build of desaputro.com, and the barrier that had held for sixteen years just collapsed. The interesting part is what happened next, because I did not stop at a homepage.
A personal site normally means a hero section and a contact form. This one did not stay that way. Because each new piece was suddenly cheap to add, it kept growing. It got a real content management system so I could write posts without touching code. It got prerendering so the blog would show up properly for search engines. It got dark mode, a house-ad system, editable URL slugs, and brute-force protection on the login. Every one of those was a "someday" item I would never have justified spending a weekend on in isolation. Lower the cost of each addition, and a small project quietly turns into a real one.
And still I did not stop. Within days I launched a second, completely different site, an e-commerce game top-up store at melogaming.com. Before AI tools, a solo developer with a full-time job launching two distinct platforms in one week would have been a fantasy. Cheaper building did not make me build less. It freed me to build far more than I set out to.
Lowering friction unlocks a backlog you forgot you had
Every developer and every company sits on a pile of ideas that never ship. Side projects left unfinished. Internal tools pushed back another quarter. Automation scripts skipped because the payoff never justified the developer hours.
When AI drops the time to build a feature from three days to an afternoon, that backlog suddenly becomes viable. Teams build the internal dashboard and the data pipeline they kept deprioritizing. Solo developers test product ideas that were never worth the upfront risk. Upgrading a legacy system becomes feasible when an agent can handle the tedious translation work. The demand was always there. What changed is that the cost finally dropped below the threshold where acting on it makes sense.
The uncomfortable half of the paradox
There is a flip side I do not want to skip. If cheaper building means we build more, we also build more junk. More half-finished side projects. More features that exist because they were easy to add, not because anyone actually needed them. Every one of those still has to be maintained, secured, and eventually understood by whoever touches it next.
I felt this on my own site. Each cheap addition was also one more thing that could break, one more surface to keep patched, one more path a bug could hide in. The content system, the ad logic, the auth hardening: adding them was an afternoon each, but owning them is forever. The Jevons boom in software is real, but it is not free. Cheaper to build has never meant cheaper to own, and the gap between those two is where a lot of this newly cheap software is quietly going to rot.
The role moves upstream
None of this means the job stays the same. AI is not eliminating developers, but it is aggressively changing what the work is. When generating a function becomes trivial, the bottleneck moves. The hard part is no longer typing the code. It is knowing what needs to exist, how the pieces fit together, and how to deliver something people actually want to use.
The developers who do well are not racing AI to write syntax faster. They are using it to move up the stack, from typist to something closer to a product builder and system architect. The scarce skill is judgment about what to build and how it should hold together, and that is precisely the part AI does not do for you.
The whole "AI will kill software engineering" argument assumed the world's demand for software was fixed, a pie that automation would let fewer people carve up. But software demand is not fixed. As long as there are problems to solve, experiences to improve, and ideas worth trying, making code easier to write only widens what is possible. We are not building less. We are just getting started.