Once upon a time, there was a small business owner who hired workers and crafters to do carpentry work: framing houses, roofing, all kinds of construction. These workers spent their days swinging hammers and driving in nails to get the job done. Then one day, the business owner discovered a new tool: the nail gun.

With this tool, the workers could drive in nails much faster and with much more precision. Productivity skyrocketed. The workers were able to accomplish their tasks quicker, with less physical strain, and they got a lot more done. The business owner was able to take on more jobs, earn more revenue, and margins soared. Everyone was happy. In fact, the business owner was so happy they gave all the workers raises.

Life was good.

Then one day, the business owner had what seemed like a brilliant idea. “This tool really helped people be way more productive,” they thought. “So really, the tool is more important than the people. I don’t need these expensive workers that I gave raises to. I’m going to replace them with people who have less experience and training. I can pay them less, but with these tools, things will still be great.”

So the business owner laid off all of their experienced workers, hired a bunch of new people with no experience or training, gave them nail guns, and sent them off to the job sites.

Quality plummeted. Work didn’t get done to the same standard. Business fell off. Eventually, all the customers went elsewhere.

This is an analogy.

Right now, I’m watching companies across the tech industry lay off their experienced, skilled engineers because “AI will replace them.” The thinking goes like this, theoretically.

AI makes development so much easier that we don’t need expensive senior talent anymore. We can hire cheaper, less experienced people, hand them AI tools, and everything will be fine.

This is the nail gun fallacy. And here’s why this fails. AI coding tools are genuinely incredible. Using Claude Code, Kiro, or Gemini, I am literally 10 to 100 times more productive as a developer. I’m writing fully featured applications in days instead of weeks or months. The productivity gains are real and they are massive.

But here’s what else I discovered. I’ve experimented with “spec-driven AI coding”. This is the approach where you just describe what you want and let the AI build it without experienced human guidance. It was a miserable failure. I tried three different apps this way. None of them ever even became functional.

The difference? When I’m driving the process with my decades of experience, AI is a force multiplier. The AI helps me with “Adjust this method to do X instead of Y” or “What’s the method signature for API X?” or “Help me find all places where this function is called”. This is me still “developing” using my knowledge, experience, and judgement, but just having the AI more quickly write the next 4-5 lines of code, which I can then quickly look at and verify. (It’s like syntax completion on steroids.) But when AI is expected to replace that human knowledge and experience entirely, you get garbage.

AI is an extraordinary productivity enhancement tool. It is not a replacement for skilled workers. The companies currently gutting their engineering teams in the name of “AI transformation” are going to learn this lesson the hard way, just like the fictional business owner with the nail guns.
The tool makes the craftsman more powerful. The tool does not replace the skill and experience of the craftsman.