Cognition CEO Calls for Output-Based Performance Reviews Over AI Token Usage Metrics This Month

According to a Founders podcast episode monitored by Beating, Cognition co-founder and CEO Scott Wu this month criticized companies that evaluate programmer performance based on AI token consumption, calling the practice "overdone." Wu recommended measuring employee performance by actual completed tasks and output instead.

The trend, known as "tokenmaxxing" in Silicon Valley, refers to employees reflexively increasing their usage of AI coding tools like Claude, Codex, or Cursor to gain advantages in internal dashboards or performance reviews. Legal AI startup Legora's technical director Jacob Lauritzen similarly criticized the practice on a podcast, stating that employees displaying token usage metrics in performance evaluations amounts to an inefficient assessment approach.