I was wrong.
Not about everything. But about enough. If you've been following me for a while, you might remember my position on AI-assisted development. I called it autocomplete on steroids. I mocked vibe coding. I was the guy in the comments pushing back every time someone posted about their AI-generated miracle app.
Then I started watching people I respected change their minds. Not the hype crowd - the serious engineers. The ones whose opinions I'd actually trusted for years. And I realized that if this was real and I refused to look, I wasn't being principled. I was being stubborn - even for me - and reacting out of fear, not reason.
I could be wrong again. Maybe we're all about to lose our jobs. But I'm betting our jobs change rather than disappear - because that's been the outcome every single time something was supposed to make developers obsolete.
What I Got Wrong
Let me be honest about what was actually happening. I was afraid. Afraid of becoming obsolete. Afraid that the thing I'd spent my entire career getting good at was about to stop mattering. When you've been writing code since you were eight years old and building software professionally since 1997, the idea that a machine might make all of that irrelevant isn't an abstract concern. It's existential.
I did what scared people do. I dismissed it. I mocked it. I found every bad example I could and held it up as proof that the whole thing was a joke. And I did all of that without actually trying the tools myself. That's the part that bothers me - not that I was wrong, but that I let the fear make me shortsighted. I had a strong opinion with no firsthand experience behind it, and I should have known better.
So I spent months - not days, not a weekend hackathon - months learning how to actually work with AI coding tools. Deliberately. Methodically. The same way I learned Objective-C in 2008, the same way I learned Swift when it dropped, the same way I've learned every tool that ended up mattering. I sat down, shut up, and did the work.
What I found surprised me. The fear was wrong - but not in the way the hype crowd thinks.
What I Got Right
I wasn't wrong about everything, though.
The people who can't write code and think AI means they don't have to learn? They're still wrong. The vibe coders who prompt their way to a demo and call it a product? Still building on sand. The LinkedIn influencers posting "I bUiLt a SaaS in FoUr HoUrS wItH AI" and neglecting to mention it falls over the second a real user touches it? Still full of shit.
AI doesn't replace the need to understand what you're building. It replaces some of the typing. Those are very different things.
The Binary Is Broken
The industry has framed this as two camps. Camp one: AI is going to replace all developers, learn prompt engineering or die. Camp two: AI is a toy, real developers don't need it, it's all hype.
Both camps are wrong, and they're wrong in the same way - they're both treating AI as a single thing that either works or doesn't. That's not how any tool works. A table saw doesn't replace a carpenter. It also isn't useless. It makes a skilled person faster. It makes an unskilled person dangerous.
I've been building software professionally since 1997. I've been writing code since I was eight years old. That experience didn't become obsolete the moment Claude learned to write a for loop. It became the thing that lets me tell Claude when it's writing the wrong for loop.
The Difference Experience Makes
I was terrified that AI would make experienced developers obsolete. The reality is the opposite. It makes us extraordinarily valuable - because we're the ones who can actually use it well.
When an AI tool generates code, someone has to evaluate it. Someone has to know whether the architecture makes sense, whether the approach will scale, whether the edge cases are handled, whether it's going to pass App Store review, whether it's the kind of code that's going to haunt you in six months. Someone has to know what good looks like.
That's not a junior developer skill. That's not a prompt engineering skill. That's a "I've shipped dozens of apps and watched hundreds of decisions play out over years" skill. It's judgment. It's taste. And you can't prompt your way into it.
This is where the fear inverts. I was terrified that AI would make experienced developers obsolete. The reality is the opposite. It makes us extraordinarily valuable - because we're the ones who can actually use it well.
The vibe coder sees code that compiles and calls it done. The experienced developer sees code that compiles and asks why it chose that approach, whether the error handling is real or decorative, and whether this is going to blow up the first time someone's phone loses connectivity in an elevator.
And here's the part that should genuinely concern this industry: what happens to the pipeline? If companies stop hiring junior developers because AI can generate code cheaper, and the juniors who do get hired never learn to actually evaluate what they're building because the AI wrote it - who becomes the senior developer in ten years? Who develops the judgment? The experience that makes AI-assisted development work isn't something you're born with. It's something you build over decades of writing bad code, shipping it, watching it break, and learning why. If we don't create junior developers with the skills and critical thinking to become senior developers, what happens when all the senior developers retire?
The Accidental Manager
Here's another part I didn't see coming.
There's a false binary in tech careers that's been around forever: grow into management or stagnate. For years the only way to advance was to stop doing the work and start managing the people who do the work. The industry has gotten better about this - staff and principal engineer roles exist now specifically because companies realized they were losing their best technical people by forcing them into org charts. But for a long time, the message was clear: manage or plateau.
I tried the management path. I hated it. I "demoted" myself back to individual contributor and never looked back. I'm a builder. That's what I'm good at, that's what I care about, and managing humans was making me worse at both.
And now, after all of that, I'm a manager again.
Except I don't manage people. I manage AI agents. I set direction, define constraints, review output, reject bad work, and push for better results. I'm doing everything a good engineering manager does - except my direct reports are LLMs, they don't have feelings about my code reviews, and they never schedule a meeting that should have been a Slack message.
It turns out the skills I hated using on people - the constant steering, the quality checks, the "No, go back and do it differently" conversations - work beautifully when your team runs on GPUs instead of coffee.
The Force Multiplier
Twenty-five years of experience didn't become worthless. It became a force multiplier.
The thing I was most afraid of - that my experience would become worthless - turned out to be exactly backwards. Twenty-five years of experience didn't become worthless. It became a force multiplier.
I can build things faster now. Not because I stopped being a programmer, but because I stopped confusing typing with building. The act of writing code was never the hard part. The hard part was always knowing what to write, why to write it, and when to stop. AI handles the first part better than I expected. It's useless at the other two.
The future isn't AI replacing developers, but it's not AI being irrelevant either. It's experienced developers who know how to wield it - who have the judgment to direct it, the knowledge to evaluate its output, and the discipline to reject it when it's wrong. Not a developer replaced by a chatbot. Not a Luddite pretending the tools don't work. It's a professional using every tool available to do better work, the same way professionals have always done.
I was a skeptic. I did the work. I changed my mind. And I'm building better and faster because of it.
If you're still in the "AI is useless" camp, I get it. I was there. But do the work before you decide. You might be surprised.
Top comments (0)