Meta has removed AI tool usage from its engineers’ performance reviews following an internal memo from executives Maher Saba and Santosh Janardhan. The company will no longer track token consumption or rely on dashboards to measure productivity, shifting focus instead to the quality, speed, and complexity of individual work. This decision follows a period where employees engaged in tokenmaxxing, burning through credits simply to improve their standing on leaderboards. The strategy proved expensive, with internal AI costs projected to reach billions by 2026. Consequently, Meta plans to introduce strict budgets and a central dashboard starting in 2027 to control spending. Simultaneously, the firm is testing its new agent tool Hatch, which aims to perform computer tasks autonomously. However, trials face resistance as staff express privacy concerns about connecting the software to personal accounts.
The policy shift signals a move away from vanity metrics toward sustainable resource management within large tech organisations. By removing the incentive for excessive token usage, Meta hopes to reduce waste while maintaining engineering output without constant surveillance. The introduction of budgets suggests a future where AI access requires financial justification rather than automatic provision. This approach may influence how other companies balance innovation costs with operational efficiency.
* Token counters and dashboards are no longer review criteria
* Internal AI spending is expected to hit billions in 2026
* New budget rules and a central dashboard launch in 2027
