The AI doesn't reduce complexity. It usually multiplies it.
In the midst of a wave of anxiety about artificial intelligence displacing software developers, one technology executive is urging the industry to distinguish between the power of a tool and the depth of a craft. Adrian Valerio Anacleto, founder of Epidata, argues that AI coding assistants accelerate output but do not dissolve the structural complexity at the heart of software development — a complexity that, if anything, deepens as digital systems grow more interconnected. He has seen this pattern before: inflated narratives about easy entry into programming swept Latin America not long ago, and while the hype was overstated, the underlying opportunity proved real. The lesson, he suggests, is that the story we tell about technology and labor is rarely as simple as the tools themselves.
- The rise of AI coding tools like GitHub Copilot has sent a tremor through developer communities, with many wondering whether their expertise is becoming obsolete in real time.
- Anacleto pushes back sharply, calling predictions of programmer extinction 'almost technological flat-earthism' — a sign that the loudest voices often understand the least about how software is actually built.
- The same ecosystem once promised that a few weeks of bootcamp training would make anyone a wealthy developer overnight, a narrative that was exaggerated yet still opened genuine career pathways for hundreds of thousands across Latin America.
- The danger now is the inverse distortion: AI tools handed to people without deep technical grounding don't simplify hard problems — they tend to multiply them, producing code that looks functional until it has to be maintained or scaled.
- As global digital infrastructure grows more complex, the demand for programmers with genuine structural knowledge of software systems is expected to rise, not fall — the tools will evolve, but the underlying craft will remain indispensable.
The question cutting through the tech industry right now is deceptively simple: if AI can write code in seconds, why learn to program at all? Tools like GitHub Copilot have made the anxiety concrete, and developers across the world have watched demos of AI-generated software and wondered if they were witnessing their own replacement.
Adrian Valerio Anacleto, who founded and leads Epidata, is not persuaded. Calling the disappearing-programmer thesis 'almost technological flat-earthism,' he argues that those making such predictions fundamentally misunderstand how software gets built at scale. For him, this is a familiar pattern. Not long ago, the same industry convinced itself that programming was a three-month sprint to wealth — some platforms promised employment in weeks, riches in months. Anacleto watched people repeat this with conviction, and others repeat it knowing it wasn't true. Yet something real emerged from the noise: hundreds of thousands of young people across Latin America found their way into software careers. The narrative was inflated; the opportunity was genuine.
Now the pendulum swings in the opposite direction, and Anacleto sees the same mechanism at work. The critical difference is that this time, the underlying claim misses something fundamental. Adding AI to the toolbox doesn't dissolve software's complexity — it often amplifies it. A developer without deep structural knowledge who deploys an AI assistant doesn't eliminate hard problems; they inherit them at greater speed and scale. Knowing why a system fails, how to make it scale, how it behaves when connected to other systems — that kind of knowledge takes years to build, and no generative tool substitutes for it.
His conclusion is unambiguous: as the world grows more dependent on software, the demand for people who truly understand how to build it will grow with it. The tools will keep changing. The skill will endure.
The question arrives with the force of a tremor through the tech industry: if artificial intelligence can now write code in seconds, why would anyone bother learning to program? The tools are real. GitHub Copilot, Claude, and their siblings have made the anxiety concrete. Thousands of developers have watched demos of AI systems spinning out functional software and wondered if they were watching their own obsolescence.
Adrian Valerio Anacleto, who founded and runs Epidata, has a blunt response to the doomsayers. Claiming that programmers will vanish because of AI, he says, is "almost technological flat-earthism." The argument sounds simple until you sit with what he's actually saying: the people making these predictions don't understand how software actually gets built at scale.
This isn't his first time watching the industry chase a false narrative. A few years back, the same ecosystem—driven by bootcamp companies, education platforms, public policy, and genuine enthusiasm about digital transformation—convinced itself that programming was a three-month sprint to wealth. Two weeks of coding lessons, some versions promised, and you'd be hired. Two months and you'd be rich. Anacleto watched people repeat this with conviction, and he watched others repeat it knowing perfectly well it wasn't true. Yet something real did happen in that cycle. Hundreds of thousands of young people across Latin America actually did find their way into software work. They built careers. They moved up. The narrative was inflated, but the opportunity was genuine.
Now the pendulum swings the other way, and Anacleto sees the same pattern emerging. The difference is that this time, the underlying claim misses something fundamental about what programming actually is. Software development, he argues, carries a weight of complexity that doesn't evaporate when you add AI to the toolbox. In fact, it often gets worse. When someone without deep technical knowledge deploys an AI coding assistant, they don't eliminate the hard problems—they amplify them. The AI doesn't reduce complexity. It usually multiplies it.
This is where the conversation shifts from hype to something more grounded. Building serious software requires understanding not just what tools can do, but how the machinery underneath actually works. You need to know why a system fails, how to make it scale, what happens when you connect it to other systems. You need structural knowledge—the kind that takes years to develop. An AI that generates code without that knowledge is like having a very fast typist who doesn't understand grammar. The output looks impressive until someone has to maintain it.
Anacleto's conclusion is straightforward: programming will remain central to the digital economy, and its importance will only grow. As the world becomes more complex, more dependent on software, the demand for people who actually understand how to build it will increase, not shrink. There will always be more things to program. The tools will change. The skill will endure.
Notable Quotes
Saying programmers will disappear because of AI is almost technological flat-earthism— Adrian Valerio Anacleto, CEO of Epidata
The AI doesn't eliminate that complexity. It usually amplifies it— Adrian Valerio Anacleto
The Hearth Conversation Another angle on the story
When you say AI amplifies complexity rather than reducing it, what do you mean exactly?
If you use an AI tool without understanding the underlying systems, you end up with code that works in isolation but breaks when it touches the real world. You've created a new problem to solve.
So the bootcamp narrative from years ago—the one about becoming a programmer in weeks—that was always wrong?
It was wrong about the timeline, but it wasn't entirely false about the opportunity. Thousands of people did enter the field and build real careers. The lie was about how fast it could happen, not whether it could happen at all.
And now we're doing the same thing with AI, just inverted?
Exactly. Instead of "anyone can code in eight weeks," it's "nobody needs to code anymore." Both miss the actual complexity of what software development is.
If AI keeps improving, won't it eventually handle that complexity?
Maybe. But right now, the more powerful the tool, the more dangerous it becomes in untrained hands. You need judgment to use power well.
So what should someone do if they're thinking about learning to program today?
Learn it the way it's always been learned—by understanding the foundations, not just the shortcuts. The tools will change. The thinking won't.