AI, debugging, and the art of staying in control

I have recently worked with ChatGPT on the clock, and the experience has been like having a new colleague—always available, dedicated, and eager to help. Like my dog, he loves me no matter what I do.

For years, debugging meant hours of frustration, only to have the real solution come to us later—maybe during a walk in the park. That Eureka moment often required time, distance, and deep thinking.

But ChatGPT has changed the game.

• Debugging is now instant—solutions come in seconds instead of hours.
• Refactoring no longer feels risky—AI provides immediate insights into whether a change improves or harms the system.

However, when changing code fast, it’s easy to lose focus on what matters most—good software architecture.

• Clean code, simple components, and loosely coupled patterns should always remain a priority.
• Reducing boilerplate and unnecessary complexity should be at the forefront of our activities.

Can moving our code too fast come with risks?

However, when changing code fast, it’s easy to lose focus on what matters most—good software architecture.

• Clean code, simple components, and loosely coupled patterns should always remain a priority.
• Reducing boilerplate and unnecessary complexity should be at the forefront of our activities.

Yet, ChatGPT does not have these spontaneous thoughts. It thinks much faster than we do, analyzing large amounts of code in seconds without mistakes. For AI, concerns about simplicity, maintainability, and future-proofing are not as relevant as they are for humans.

My experience has been to prevent the machine from thinking ahead of me. The goal is always to extract AI’s insights while ensuring the codebase remains clean and understandable—so that at any time, I could work on it alone if needed.

This raises an important question:

• Have I developed this mindset simply to protect myself from becoming too reliant on AI?
• Or is this the right approach to ensure that the systems I build remain scalable and maintainable in the long run?

This article is an attempt to answer that question.