Will There Ever Be a 10x Prompt Engineer?

The best experience I’ve had with AI code generation is when I’m “piloting the plane,” so to speak, occasionally putting it on autopilot, but never leaving the cockpit. Always monitoring, ready to take over when needed.

Writing code with AI assistance will 10x your productivity. That’s what everyone says. And sure, in many ways they’re right. But there are two hard parts nobody mentions. First is losing track of where the code is headed. Second is knowing when to step in and correct course. That second one is critical, and it’s what junior developers today might never develop.

I feel fortunate to have a decade of professional experience without AI. Some might argue I’m the unlucky one. Maybe. But I don’t think so.

The gap between great developers and average ones has always been significant. With AI, I see it widening. The best get better because they know exactly when to use the tool and when to take control. The average stays about the same, maybe slightly more productive. The worst get worse, producing more code they don’t understand at a faster rate.

But prompt engineers? That skillset gap is narrow. Can you imagine a prompt engineer being so incredibly skilled, so good at prompting that they become rare and uniquely valuable? I can’t. Every time someone discovers a clever prompting trick, it’s either built into the next model or shared in a viral post within days. The knowledge spreads instantly.

Coding is learned through experience. Through late night firefights…that you caused yourself. Real people being impacted by your bugs. Through expanding legacy crap code that makes you question everything. And when you finally get your chance to refactor that garbage, you replace it with whatever trendy pattern is hot that year. Just like your predecessor did. Then you watch the cycle repeat.

But after a few cycles, you learn. Not syntax or patterns. Judgment. When to refactor and when to leave it alone. When that clever abstraction will bite you in six months. When boring code is better than smart code. This is how developers earn their stripes.

I feel bad for junior devs starting out today. That may seem counterintuitive, since AI does make it easier than ever to build, and anyone can launch apps with enough AI tools. It’ll only get easier too.

But imagine trying to learn when you don’t know enough to recognize when the generated code is subtly wrong. When it’s elegantly solving the wrong problem. And when it’s adding complexity that will haunt the codebase for years.

What happens to the next generation of developers? How do they earn their stripes when AI handles the implementation? Maybe software engineering is forever changed into something completely different. But I don’t see a world where prompt engineers bring lasting value. The best engineers will always be significantly better than average ones. That gap represents real value. Years of accumulated judgment and experience. The skill isn’t knowing how to prompt. It’s knowing what to build, when to build it, and most importantly, when not to build it at all.

If you’re starting out in this AI world, be self-aware. Learn to fly the plane before relying on autopilot. Use AI chat to ask all your dumb questions. Kill that imposter syndrome. Keep a misc project for wild experiments. Go crazy there. See what happens when you let AI completely drive. Break things where it doesn’t matter.

But on real work? Limit the code generation. Don’t let AI do the work. Use it to explain concepts. Force yourself to type it out even when AI could do it faster. You should be optimizing for learning and not speed.

The autopilot is incredible, but only if you know how to fly when it fails. And it will fail. Not obviously. Not loudly. Subtly, in ways only experience will help you recognize.

The question isn’t whether AI will change software development. It already has. The question is whether the next generation will still know how to fly when the autopilot breaks.