DC Studio | shutterstock.com
It seems that software developers at Meta who are fully committed to AI-powered programming work have found a new way to measure their engagement. This manifests itself in the form of an internal ranking that records which dev uses the most tokens with Claude Code. This is also called “Tokenmaxxing” and is one of the stupidest ideas I have come across in a long time.
Because in this race to see who uses the most tokens, it doesn’t matter whether it resulted in a good result. The rankings are simply there so that those involved can adorn themselves with dubious titles like “Token Legend” or “Cache Wizard”. Something similar is apparently happening at Microsoft, OpenAI and Salesforce.
3 reasons against tokenmaxxing
Managing software developers is actually hard enough. There are various reasons for this, but the most important one is that the software development process and the productivity of the developers are very difficult (or impossible) to measure. Which isn’t to say that it hasn’t been tried. On the contrary: Over the years, lines of code, story points, total working hours and per task, and the weekly bugs fixed were measured – among other things. All of these “key figures” have two things in common: they don’t work and in the end they are just manipulated.
