AI Will Write 90% of Code—and We'll Still Need More Developers
When Anthropic's CEO predicted that AI will be writing 90% of all code within a year, a lot of people took it as a sign that programming is over. The end of human developers, the rise of machines, the usual sci-fi panic. We've seen this fear every time a tool threatens to upend the way work gets done, from the industrial loom to the desktop computer.
This time really is different. But not in the way people think.
We're about to see an explosion of software. Far more than we've ever had, filling niches that today seem too small, too weird, or too unprofitable to bother with. AI won't just replace code that people are writing now—it will unlock an entire universe of software no one's been able to afford.
All the Software That Doesn't Exist (Yet)
Everywhere you look, there's software missing from the world. Small businesses, side hustlers, volunteer groups, schools, solo entrepreneurs—they all have ideas for apps or tools that would make their work and lives easier. Most of them stay stuck on paper because hiring developers is expensive, and the off-the-shelf solutions are bloated, overpriced, or just don't fit.
The neighborhood soccer league that wants a clean way to track game stats. The community theater that could really use a simple ticketing system. The freelance designer who's tired of wrestling with huge project management tools when all they need is a basic task board.
- A meditation app built around one specific practice, instead of a mashup of 50.
- A personal finance tracker that doesn't assume you have a hedge fund.
- A lightweight app to help coworkers coordinate hybrid schedules—because the existing tools try to be too much.
Right now, these ideas get shut down before they start. Too expensive. Too custom. Too small a market. They can't compete with enterprise-scale services, and it's not worth paying for bespoke builds. So they either hack together messy solutions or give up entirely.
The Cost of Building Just Collapsed
AI is crushing the cost of making software. What once took a team of developers weeks to deliver can now be scaffolded in hours. Suddenly, it's not just the big players who can afford custom tools. It's everyone.
Whenever the cost of creating something falls off a cliff, demand doesn't shrink—it explodes. This is Jevons' paradox in action: the more efficient we get at producing something, the more of it we end up using overall. Software is no exception. AI makes building software dramatically cheaper and faster, which means way more people will want way more of it.
We've seen this play out in photography, publishing, video production. The cheaper and easier it is to create, the more people do it. Not just more of the same, but entirely new kinds of projects that never made sense before.
The Age of "Good Enough" Software
We are moving into an era of good-enough software. Apps that don't aim to be perfect or flashy but are perfectly adequate to do exactly what someone needs—and crucially, cheap enough to exist in the first place.
- A mini CRM that keeps a small business organized, even if it's not dripping with features.
- A booking app that handles appointments with zero upsell for "advanced marketing tools."
- A dead-simple dashboard for a warehouse that just needs to count things and send reports.
None of these would win design awards or show up in glossy product roundups. But they'll be invaluable to the people who need them. And AI is making them viable where before there was nothing at all.
Everyone Becomes a Software Builder
The fascinating thing about Jevons' paradox is how it reshapes the entire landscape. As software gets cheaper and easier to build, we don't end up with fewer developers—we end up with more people building software, period. Developers, designers, product managers, even people who'd never have dreamed of making an app before. Everyone starts playing a role, often using AI as a co-pilot or assistant to translate ideas into code.
The line between "developer" and "user" blurs. Some people will still write intricate code by hand; others will describe what they want in natural language and refine AI's output. In both cases, the work of making software expands, even as the manual labor shrinks.
The Frontier Stays Human (for Now)
Of course, there's still a frontier AI can't cross yet. Critical systems, high-performance infrastructure, totally new architectures and inventions—these don't build themselves. That's where human expertise keeps pushing forward, solving problems AI doesn't know how to tackle and inventing the next generation of tools.
More Software, More Builders, More Weirdness
AI really might end up writing the vast majority of code. But that doesn't shrink the opportunity space. It blows it wide open. All the tiny, overlooked, niche, or too-expensive ideas can finally get built. And as that happens, demand for people who can navigate, supervise, and innovate will keep growing.
It won't just be developers. It'll be everyone figuring out how to turn their ideas into tools—because the gap between idea and software is about to get vanishingly small.
What's coming isn't the end of software development. It's the start of something weirder, more chaotic, and a lot more interesting.