Another Broken Thing We Won’t Fix: Measuring Engineer
AI made the numbers meaningless. So play the game everyone’s already playing, and stop feeling bad about it.
Somewhere right now a company is deciding how good an engineer is by counting the AI tokens they burn. More tokens, more productive. I’m not making that up, teams actually did this in 2026, there’s a name for gaming it, tokenmaxxing, people running up their usage so the number looks busy. And honestly it’s the perfect picture of where we’ve landed. The old way of measuring engineers quit working, so we reached out and grabbed the first countable thing we could touch, and it was about the dumbest one in the room.
Everyone already knows the old numbers died. Lines of code, PRs per week, story points, velocity. AI made all of it go straight up without the work getting any better, and if we’re honest the numbers were shaky long before that. So, fine. That part’s done, it got written, everyone nodded, we moved on. There’s a whole pile of these now, things AI quietly broke that nobody has an honest fix for, and how we measure engineers just got added to the pile. The part I can’t stop looking at is what came right after. The scramble to find the number that replaces it.
