Amazon has shut down an employee-built internal leaderboard called KiroRank, which ranked staff by their usage of Kiro, the company’s in-house AI coding tool, and awarded PhoneTool achievement badges to top users. The Financial Times first reported the closure; Amazon’s official internal announcement framed it as mission accomplished, saying the program had created “awareness about what AI can do” and that adoption had reached its goals. Multiple Amazon employees told 404 Media they believe the actual driver was that the leaderboard was trivially easy to cheat and was pushing employees toward expensive, unproductive AI consumption. One employee said they began gaming the system directly after being told in a performance review that they were not using AI enough — describing it as effortless to automate an endless series of tasks unrelated to their real work. “Honestly, iterating on that and maximizing the throughput was the most fun I’ve had at work,” they said. “I also do not think I was the only one gaming the system to make the number go up. My manager’s tone in that meeting made me think there were some internal discussions about the program driving waste.”
Amazon told 404 Media that KiroRank “was not a formal or approved tool” but rather a beta dashboard created by a group of employees, and that it does not mandate AI tool usage or track employee usage — though it does measure token utilization to understand cost and efficiency patterns. The episode is a textbook example of what the article calls “tokenmaxxing”: the management philosophy, now common across the tech industry, that employees not maximizing AI usage metrics are underperforming. The practical result has been workers running background scripts to inflate AI usage numbers without doing anything productive, burning compute costs for appearances while actual productivity gains remain unmeasured. One Amazon employee who genuinely finds Kiro useful observed that the leaderboard had created incentives “to not bother trying to be efficient on token use” — rewarding wasteful over-prompting rather than thoughtful adoption.