Read
No, AI is not Making Engineers 10x as Productive
10x productivity means ten times the outcomes, not ten times the lines of code. This means what you used to ship in a quarter you now ship in a week and a half.
Emphasis mine.
And also:
I think a lot of the more genuine 10x AI hype is coming from people who are simply in the honeymoon phase or haven’t sat down to actually consider what 10x improvement means mathematically. I wouldn’t be surprised to learn AI helps many engineers do certain tasks 20-50% faster, but the nature of software bottlenecks mean this doesn’t translate to a 20% productivity increase and certainly not a 10x increase.
I do think that we are all writing more software than we used to—I know I am writing little scripts and applications that I never would have taken the time to write without AI. They make me more productive, in a way, but they are also often taking the place of other things I would have spent equal time on: a script instead of a daisy-chained automation, a microsite communicating data interactively instead of a Google doc doing the same in a static manner, a prototype to communicate a design instead of a static mock.
But—and this is the big qualifier—“throw away” (AI-written, used once-ish) software isn’t necessarily contributing to the bottom line of my production code. Throwing more lines at production-level work for the sake of more lines is a recipe for disaster. I still need to know what that code does and I still need it written at sensible pace that I can understand. I often use AI to help write it, but this code is treated very differently.
I suspect we’ll have at least 2 tiers of software in the future: software AI wrote for us that accomplishes a task and is either never used again (or not distributed in a way that bears the meaningful burden of support), and software that is production-ready. The latter may still be increasingly LLM-written, but I don’t think we’ll want to be flippant about the code that we allow LLMs to contribute to production-level software.