Maintain Your Instinct Sharp Whereas Utilizing AI for Coding

This article is crossposted from IEEE Spectrum’s careers newsletter. Sign up now to get insider tips, expert advice, and practical strategies, written in partnership with tech career development company Parsity and delivered to your inbox for free!

How to Keep Your Engineering Skills Sharp in an AI World

Engineers today are caught in a strange new reality. We’re expected to move faster than ever using AI tools for coding, analysis, documentation, and design. At the same time, there’s a growing worry in the background: If the AI is doing the work, what happens to my skills?

That concern isn’t just philosophical. Research from Anthropicthe company behind Claude, has suggested that heavy AI assistance can interfere with human learning—especially for more junior software engineers. When a tool fills in the gaps too quickly, you may deliver working output without ever building a strong mental model of what’s happening underneath.

More experienced engineers often feel a different version of this anxiety: a fear that they might slowly lose the hard-earned intuition that made them effective in the first place.

In some ways, this isn’t new. We’ve always borrowed solutions from textbooks, colleagues, forums, and code snippets from strangers on the internet. The difference now is speed and scale. AI can generate pages of plausible solutions in seconds. It’s never been easier to produce work you don’t fully understand.

I recently felt this firsthand when I joined a new team and had to work in a codebase and language I’d never used before. With AI tools, I was able to become productive almost immediately. I could describe a small change I wanted, get back something that matched the existing patterns, and ship improvements within days. That kind of ramp-up speed is incredible and, increasingly, expected.

But I also noticed how easy it would have been to stop at “it works.”

Instead, I made a conscious decision to use AI not just to generate solutions, but to deepen my understanding. After getting a working change, I’d ask the AI to walk me through the code step by step. Why was this pattern used? What would break if I removed this abstraction? Is this idiomatic for this language, or just one possible approach?

The shift from generation to interrogation made a massive difference.

One of the most powerful techniques I used was explaining things back in my own words. I’d summarize how I thought a part of the system worked or how this language handled certain concepts, then ask the AI to point out gaps or mistakes. That process forced me to form my own mental models rather than just recognizing patterns. Over time, I started to build intuition for the language’s quirks, common pitfalls, and design style. This kind of understanding helps you debug and design, not just copy and paste.

This is the core mindset shift engineers need in the AI era: Use AI to accelerate learning, not to replace thinking.

The worst way to use these tools is also the easiest: prompt, accept, ship, repeat. That path leads to shallow knowledge and growing dependence. The better path is slightly slower but more durable. Let AI help you move quickly, but always come back and ask, Do I understand what I just built? If not, use the same tool to help you understand it.

AI can absolutely make us faster. Used well, it can also make us better at our jobs. The engineers who stay sharp won’t be the ones who avoid AI, they’ll be the ones who turn it into a collaborator in their own learning.

—Brian

When war strikes, critical power infrastructure is often hit. Engineers in Ukraine have risked their lives to keep electricity flowing, and some have been hurt or killed in the dangerous wartime conditions. One such engineer, Oleksiy Brecht, died on the job in January. “Brecht’s life and death are a window into the realities of thousands of Ukrainian engineers who face conditions beyond what most engineers could imagine,” writes IEEE Spectrum contributing editor Peter Fairley.