Your AI Usage Is Not Your Output. My Biggest Claude Month Was My Quietest on GitHub.

Amazon says token counts measure the wrong thing. My August cost $4,950 at list price and logged 99 GitHub contributions.

A tachometer measures how hard the engine works, not how far the car goes. Photo by Matthias Speicher on Unsplash; text added by the author.

Amazon put it in writing on Tuesday.

A senior engineer told its developers that counting their AI use is a “deeply flawed” way to judge their work.

Accenture, two weeks earlier, capped many of its staff at $29 a month of Claude.

So I counted mine. My most expensive AI month was my quietest on GitHub.

The memo is on Amazon’s own site, signed by Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business. It names the numbers managers have reached for since coding assistants arrived: “token usage or adoption of AI tools”, and “code changes and reviews”. Then it says, “We are all aware of how deeply flawed these metrics are, primarily because they are disconnected from what matters.” CNBC reported that staff first saw it in September. Fortune reported that Amazon had already shut an unofficial leaderboard of AI use, Kirorank, in May, after employees were reportedly “running AI agents on meaningless tasks to inflate their token count”. Meanwhile, 404 Media reported that some…