I opened LinkedIn this morning and it was the same show. Half my feed is people going all-in on AI. The other half is mocking them, sharing screenshots of hallucinated code, writing threads about how AI output is full of bugs, how we still need to understand the code.
They’re right. AI-generated code today is far from perfect. Nobody serious claims otherwise.
But they’re missing the point.
The entire debate is stuck in the present tense. “AI code is buggy.” Okay, sure, today. But will it be in five years? Three? The rate of improvement is not subtle. Every few months the bar moves noticeably. If you judge AI by what it produces right now, you’re evaluating the trajectory by staring at a single frame.
The question worth asking isn’t “is AI code good enough today?” It’s “is the gap closing, and how fast?” And if you’re paying any attention at all, the answer is obvious.
Now here’s what that means. If generated code is getting good (and it is), the skills that matter are shifting under our feet. The value isn’t in typing code line by line anymore. It’s in knowing what to build. It’s in system thinking, architecture, product sense. It’s in orchestrating AI effectively, asking the right questions, reviewing output critically, iterating fast. It’s in the ability to learn and adapt when the ground moves.
The people clinging to “I write every line by hand” as a badge of honor are optimizing for a game that’s changing. The skills that brought you here are not the same ones that will take you forward.
I’m betting on the future. And from where I’m sitting, it’s getting closer faster than most people think.