There's a lot of noise about AI replacing developers. Most of it misses the point. After using AI tools daily on real production projects, here's what actually changed — and what didn't.
What AI is actually good at
The honest answer: the parts of the job that were never really the job. Boilerplate, repetitive CRUD, writing tests for logic you already understand, translating a design into markup. These tasks weren't where the value was. They were just in the way.
- Scaffolding components and API routes in seconds
- Writing tests once the logic is already clear
- Refactoring without the tedium
- Unblocking yourself on unfamiliar APIs or libraries
What still needs you
Architecture decisions, system design, understanding what the product actually needs — none of that gets easier with AI. If anything, it gets more important. When you can execute faster, the quality of your thinking becomes the bottleneck. The engineers who are struggling aren't being replaced by AI. They're being exposed by it.
How I use it day to day
I treat AI like a powerful tool that needs a skilled hand behind it. I wouldn't point it at a vague problem and hope for the best. I give clear context, review what comes back, and own the result. The output is only as good as your ability to evaluate it — which means you still need to know what good looks like.
- Write a rough implementation first, then use AI to clean it up
- Use it to explore options quickly, not to make decisions
- Always read the output — AI-generated bugs are still your bugs
The real shift
The developers who are thriving are the ones who've always been good at thinking clearly about problems. AI just amplified the gap. If you were great at understanding systems and communicating intent, you're now significantly faster. If you were mostly good at memorising syntax, you're in a harder spot. That's the real story.