This is a special post for quick takes by RohanS. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Here's a question I think feels promising for measuring the overall value generated by AI uplift relatively accurately compared to my previous vague notions of how one could go about that.
"Consider the past six months of your work and where most of the value of that work was generated. How long would it have taken you without AI?"
The past 6 months isn't necessarily the right 6 months to look at. You could also consider 6 months from 5 years ago, when AI was mostly useless, and then you might want to ask: how long would it have taken you with current AI?