The Narrative is Broken

The conversation around AI is broken because people keep arguing between two extremes: “AI is useless” v/s “AI will replace software engineers.” Neither matches what I see in practice.

My take after using AI coding agents extensively:

They’re useful. I use them. They save me time every day. But not because they’re software engineers. Because they’re fast. They can search a codebase faster than I can, read documentation faster than I can, find references faster than I can, apply repetitive changes faster than I can. You get my point. That’s valuable.

What I don’t see is evidence that they understand software the way experienced engineers do. If they did, we wouldn’t spend so much time writing architecture docs, coding standards, workflows, repository rules, skills files, and review processes just to keep them on track. The more I use these systems, the more I think their biggest strength is accelerating existing expertise, not replacing it.

Writing code was never the hard part. Understanding what should be built, why it should be built, how it fits into a system, and how to maintain it years later. That’s where the real work begins. The problem is that many non-technical people see code generation and assume engineering has been automated. A coding agent doesn’t replace engineering judgment. Those are not the same thing.

---

What concerns me more is the gap between the technology and the narrative around it.

Around the world, hundreds of billions of dollars are being committed to AI infrastructure, data centers, chips, power systems, and compute capacity. Some projections estimate trillions of dollars of AI-related capital expenditure over the coming years. Maybe those investments will pay off. Maybe they won’t. What I find strange is how often usefulness gets confused with inevitability.

A tool can be valuable without replacing expertise. A tool can improve productivity without eliminating professions. A tool can be important without becoming the center of the universe. AI is a powerful tool. A tool becoming faster is not the same thing as expertise becoming obsolete. Humanity still has enormous challenges to solve in energy, healthcare, manufacturing, education, housing, and infrastructure. AI is one technology among many, not the answer to everything.

I think it’s useful. I think it’s “Overhyped”.

And I think many people who don’t actually build software are confusing code generation with software engineering.

Those are not the same thing.