— Andrej Karpathy, 2025The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between.
AI will turn 10x engineers into 100x engineers. I've heard that statement and its variations many times over the last few months. But I have two issues with that.
Why is the multiplier constant?
I understand that "10x engineer" is just a label for a high-performing, distinguished engineer. But 100x? Sorry, but that doesn't make any sense. It obviously should depend on time. Coding LLMs are developing super quickly, so an engineer's performance is going to be C×E(t), where E(t) is a common growing efficiency function for all engineers, and C is a personal efficiency coefficient — which could be 0.1, 100, or any other real number. This now depends not only on traditional skills, but more than ever on the speed of adoption of new technologies, especially AI.
Does it even matter?
The obvious trend is that we're rapidly shifting from software engineering to direct building. And I think a great product manager with mid-level technical skills is going to be more efficient at building a real product than an old-school outstanding software engineer who is focused on code and architecture alone. And I think this could happen after just the next few releases from the AI giants already.
What do you think?