What Do We Actually Do?
A developer's job is not writing code. It never was. Writing code is the visible part. It is the part you can measure, the part that shows up in your GitHub graph. The real job is different. It is understanding a problem. It is breaking that problem into small parts. It is deciding which parts you actually need. It is seeing where the system will break before it breaks. Code is just the last step of this thinking process. It is a translation step.
A normal day looks like this. You spend three hours finding a bug. The fix takes two minutes. You think about a new feature, and you compare three different approaches in your head. Which one will cause problems in six months? In a meeting, you explain the technical side of a product decision. In a code review, you look at a teammate's pull request and ask why they built it this way. The lines of code you write are a small part of your actual time.
AI took the cheapest part of the job
When AI started writing code, it took the easiest part first. A function skeleton, a simple CRUD endpoint, a basic test file, these things now appear in seconds. This is a real gain. I don't deny it. I use it every day.
But the latest Stack Overflow survey shows something interesting. Most developers now use AI tools daily. But the top complaint is this: the output is almost right, but not quite. The second complaint is close behind: debugging AI-generated code often takes longer than writing it from scratch. So the real work, the part where you ask "is this correct, and if not, why not", still belongs to us. AI took the writing. It did not take the deciding.
The work moved from writing to checking
Here is the shift I notice in my own work. In the past, I spent most of my time producing code line by line. Now I spend most of my time reading code, questioning it, and asking why a certain choice was made. I moved from writer to editor. This is not always the easier job. Sometimes writing something from scratch is easier than following someone else's logic (or a model's logic) and finding the mistake in it.
This shift also widened the gap between junior and senior developers. Junior developers used to learn by doing simple, repetitive tasks. Now AI does most of those tasks. So the learning path for junior developers is getting narrower. For senior developers, the opposite happened. One senior developer can now do what used to take a small team, because they hand off the mechanical part to AI and spend their own time on decisions.
What hasn't changed
So back to the question. A developer's job was never "writing code." It was always "making the right call." AI did not change this. It just made it visible. In the past, writing code and making decisions were mixed together in the same act. It was hard to separate them. Now they are separate. And once you separate them, it becomes clear which one was actually valuable.
Sources
Comments ()