The Pulse: ‘Tokenmaxxing’ as a weird new trend

A growing trend called 'tokenmaxxing' has emerged at large tech companies like Meta, Microsoft, and Salesforce, where developers deliberately burn excessive AI tokens to inflate usage metrics and appear 'AI-native.' At Meta, an internal leaderboard called 'Claudeonomics' tracked token usage across 85,000 employees, resulting in 60.2 trillion tokens burned in 30 days — potentially worth $900M at retail API prices. The leaderboard was taken down after media coverage. Microsoft and Salesforce have similar dashboards with minimum spend targets, leading engineers to ask AI pointless questions or prototype features they never intend to ship. Shopify's earlier approach is highlighted as a better model: a usage dashboard with circuit breakers for runaway agents and manual review of top spenders. The author draws a parallel to the discredited lines-of-code productivity metric, arguing that token count is equally gameable and that tokenmaxxing generates massive costs with little business value.