AI and the productivity trap

Honestly, we've all fallen for it. We're living in the age of AI, and the temptation to let it do the heavy lifting is huge. The problem is, if we don't master the basics anymore, we risk ending up with apps that look great on the surface but are riddled with hidden problems. Silent bugs, invisible technical debt... until the day everything collapses.

The Tech Lead's observation

As a team lead, I see this every day. My developers are fully embracing tools like ChatGPT or Gemini, and I can't blame them—it's so much faster. But when I do a code review, I discover a bunch of stuff that has no business being there:

• Dead code hanging around for no good reason.
• Hardcoded values everywhere that should have been variables.
• Shaky or completely outdated logic.
• Old legacy code accidentally brought back to life.
• Features that mysteriously disappear between two versions.

The pace is insane, we're churning out projects, but no one is taking the time to actually read the code the AI generated. And that's where technical debt quietly piles up.

My strategy for regaining control

I realized we had to take back the reins. Here's what's working for us:

• Backend is sacred. Backend code must be checked and re-checked—no exceptions. It's the foundation of everything.
• We do TDD by hand. We don't let the AI write our tests. We write our tests ourselves, and only then do we ask the AI to write the code. We meticulously review every unit test it proposes, because we can't blindly trust it.
• Smart prompts. We force the AI to keep things simple. We ask it to apply the KISS (Keep It Simple, Stupid) and YAGNI (You Aren't Gonna Need It) principles. We don't want a massive, convoluted mess.
• We test seriously. Every feature has to be tested from A to Z. The hidden bugs are the most insidious ones.

My pro tip: I turn my prompts into to-do lists. I ask the AI to validate each step, and I verify everything point by point. It sounds simple, but it massively reduces production errors.

AI is a rocket ship, for sure, but it will never replace our critical eye and our expertise. It's up to us to learn to use it smartly, without getting lazy
.

Fassinou Claude, Developer • Tech Lead • IT Project Manager in training