Engineering video
0:56 · 16 August 2025
The speaker explains: studies show AI tools provide a real productivity increase of 0 to 30% depending on task complexity, codebase size, and language, contrasting with unreliable self-reported gains or metrics based solely on lines of code.
Hear it at 0:00 →what they found out is that on average on your typical software engineering task, AI will make you around 10 to 15% more productive. What you'll see a lot in these vibe coding communities is that they all talk about the app they build in two days, but those apps have no users and they just basically copied or cloned some existing app. That's why they are dead fast. And when they talk about productivity, they measure it as either self-reported as, oh, I feel 10 times faster, which is not reliable.
or lines of code. Like, oh, I made all these commits with AI. I would have never done this without AI. And that's okay, but when we talk about commercial software, stuff people pay for, that's way far away. Now, what they saw is a real productivity increase of 0 to 30% depending on the task and the complexity of the task, the size of the code base and the language. But what they also observe is that whenever you're using AI, you spend a lot of time reworking code.
