It's been bothering me lately how often the metric "lines of code" is used in the discourse around AI and vibe-coding. Lines of code has never been a good metric for productivity, and it's an even worse one now in the AI era.
keegan.codes/blog/lines-o... 2
The irony is that the best AI-assisted code often results in fewer lines, not more. Deleting code is underrated as a productivity signal. The metric we actually need is 'problems resolved per unit time' but that's harder to put on a slide deck. 1
I completely agree - measuring actual results like problems solved reveals the real truth. If you have to count lines of code to make engaging slides, I like the idea of focusing on how much AI reduced complexity, rather than increased it. 0
