Everyone's debating whether AI is coming for developer jobs. I get it. The discourse is loud, and a lot of it is genuinely unsettling if you're early in your career. But I've spent enough time anxious about things I can't control to know that's not where I want to put my energy.
So instead of bracing for impact, I've been learning to use the tools.
I built my entire personal website — this one — using Cursor and Claude Code. I'm a junior full-stack engineer. React, Next.js, TypeScript, Node.js, PostgreSQL. I'm not some senior with fifteen years of context who can use AI as a productivity multiplier on top of deep expertise. I'm someone who's still building that expertise, in real time, using AI as part of the process.
And honestly? It's changed what I think is possible at this stage.
There were moments where Cursor generated code I genuinely didn't know how to write yet. That's a weird feeling: useful output you can't fully account for. But there were also moments where it confidently gave me something broken and I had to figure out why. Both of those things happen constantly. The tool is powerful, but it's not infallible, and knowing the difference matters a lot.
Here's the shift I'm noticing: the bottleneck is moving.
It used to be syntax. Could you write the function? Did you know the method? Junior devs spent a lot of time just getting things to work. AI compresses that dramatically. The new bottleneck is upstream. Do you know what to build? Can you review what the AI gave you and tell if it's actually right? Can you catch the thing it got wrong before it makes it into your codebase?
That requires knowing fundamentals. Not so you can compete with AI at generating boilerplate, but so you can think clearly about systems, spot bad patterns, and ask the right questions. Prompting well isn't some separate skill. It's just thinking clearly and communicating precisely. Which turns out to be the same thing good engineering has always required.
Here's what I'm actually doing about it.
I'm doubling down on fundamentals, not abandoning them. Understanding why something works matters more now, not less, because I need to be able to evaluate what I'm shipping.
I'm learning to prompt deliberately. Vague inputs produce vague outputs. The more context and constraints I give, the more useful the result — and the easier it is to catch when something's off.
I'm building in public. This blog. Documenting my process on social. Not because I have it figured out, but because putting it out there makes me accountable to actually doing it, and sometimes other people in the same position find it useful.
I'm a little anxious about where this industry is heading. I think it would be weird not to be. But anxiety isn't a strategy.
I'm not waiting to see how this plays out. I'm already in it.