Recently, I’ve been thinking about all the commotion surrounding AI taking over jobs. Sam Altman said AI will take 95% of spokesperson roles, and then someone from Anthropic said AI could fully replace software engineers within 6–12 months. This pessimistic feeling is widespread, but I think many people misunderstand how this technology actually works.

Naval Ravikant from AngelList has an interesting take on this. He says the productivity boost from AI might be overhyped, and if we look deeper, AI always has weaknesses. No matter how advanced it is, it will still make mistakes. That’s why software engineers who truly understand the underlying logic are still needed so badly.

My view is that people who are real experts in their field don’t need to be afraid. Software engineers who understand the mechanisms beneath all these abstractions have a huge advantage. When AI generates code for you—using Claude Code or whatever—it will produce bugs and imperfect architecture. People who understand the basic logic can quickly plug those gaps when they show up. So if you want to build applications with a solid structure, high performance, and catch errors early, you still need a background in software engineering.

In many ways, traditional software engineers can leverage AI tools better than others. There are also many problems in the software world that AI can’t solve, especially when they fall outside the data it was trained on. For example, if you want to write code with extremely high performance on a new architecture—or something truly new—you still need to be directly involved.

What I’ve noticed in the market right now is that everyone wants the best. A better application almost certainly wins 100% market share in its category. The prize for first place is a Cadillac; for second, a set of steak knives; and for third? It’s gone. So the reality is: if you want to succeed, you have to be the best in some field. But that field can be anything.

What matters is this—keep redefining what you do until you find a subfield that fits you and then become outstanding there. This principle is still relevant in the AI era. Expertise and depth of knowledge in a domain still can’t be replaced. As long as you master your field, you’re safe.