AI Tools Are Power Tools. You're Still the Carpenter.

A nail gun doesn't make someone a carpenter. It makes a good carpenter 10x faster — and an amateur dangerous at scale.

That's the most honest analogy I've found for AI coding tools, and after 8+ years of shipping production software — the last few building with LLMs daily — I keep coming back to it.

The tool generates. You decide.

When people ask whether AI is replacing engineers, they're asking the wrong question. The nail gun didn't replace carpenters. It changed what carpentry is.

The judgment about what to build, whether the joint is sound, and what "done" actually looks like — that still lives entirely with the human holding the tool. Claude or Copilot can produce code faster than I can type it. Neither can tell you whether that code should exist, whether it fits the architecture, or whether it will hold up under production load six months from now.

The job moved up a level

My work didn't disappear. It shifted:

• From writing every line → to reviewing and directing. I spend far more time reading generated code critically than I ever spent typing.
• From typing speed → to judgment speed. The bottleneck is no longer how fast you can produce code. It's how fast you can evaluate whether it's right.
• From "can I build this?" → to "should it be built this way?" Capability is cheap now. Taste and architectural judgment are not.

We've seen this movie before

Nobody writes assembly anymore. Compilers didn't make programmers obsolete — they moved the abstraction level up, and the profession got bigger, not smaller.

AI is the next abstraction jump: from writing code to specifying intent and verifying output. The syntax layer is being automated. The intent layer — understanding the problem, the users, the constraints, the failure modes — is more valuable than ever.

The two failure modes

The engineers who struggle in this era fall into two camps:

1. The resisters — who treat the tools as a threat and refuse to build fluency with them. They're the carpenters insisting on hand-driving every nail while the crew next door finishes the frame.
2. The blind trusters — who ship whatever the model produces without reading it. They're the amateurs with nail guns: fast, confident, and building things that won't survive an inspection.

The engineers who win sit in the middle: fluent with the tools, and fully accountable for the outcome.

The line I keep coming back to

The nail gun is fast. The carpenter is responsible.

That's the whole deal. Use the tools aggressively. Own the output completely. Everything else about working with AI is a footnote to those two sentences.

---

I write about engineering, AI, and building software at nareshpahariya.me/insights.