What started as a quiet, incentive-driven culture in Silicon Valley is beginning to take on a more defined shape inside Meta—one where how employees use AI is becoming just as visible as what they produce.
According to a report by The Information, CEO Mark Zuckerberg recently issued what an internal memo described as a “bold ask”: engineering teams should begin rewriting Meta’s codebases so AI agents can more easily understand and modify them, effectively structuring work around AI systems.
“It’s well known that this is a priority and we’re focused on using AI to help employees with their day-to-day work,” a Meta spokesperson told The Information.
But inside the company, this new concept is evolving into something more measurable.
According to The Information, employees are increasingly tracking not just output, but interaction—specifically, how many tokens they consume while working with AI tools. Tokens, which represent chunks of text or data processed by AI systems, effectively measure how much “work” is being done between humans and machines. And at Meta, higher usage is beginning to carry its own kind of weight.
The report says the company maintains an internal dashboard that logs token consumption across roles, with software engineers leading usage but visibility extending to other teams. Over a recent 30-day period, total activity reportedly surpassed 60 trillion tokens, according to people familiar with the system who spoke to The Information.
Internally, token usage has taken on a gamified layer. According to The Information, employees compete for titles like “Session Immortal” and “Token Legend,” earning badges ranging from bronze to emerald, alongside achievement-style labels such as “Model Connoisseur” and “Cache Wizard.” A leaderboard, dubbed “Claudeonomics”—a nod to Anthropic’s Claude models—ranks the top 250 users out of more than 85,000 employees.
“The rankings…measure how many tokens employees are burning through,” one person familiar with the dashboard told The Information.
What began as a way to encourage AI adoption is now subtly influencing how work gets done, according to The Information. Some employees reportedly run extended AI sessions or assign agents to perform long research tasks—not strictly out of necessity, but to increase their visible engagement.
“You don’t want to be the one who solved it in two prompts if everyone else is showing ten,” a person familiar with the trend told The Information.
While token usage is reshaping how employees work, it also carries real-world cost implications. AI providers typically charge based on input and output tokens, with output often priced higher.
Using public pricing from Anthropic’s Claude models—around $15 per million tokens—as a rough benchmark, Meta’s reported 60 trillion monthly tokens could translate to roughly $900 million in usage. The actual figure may differ depending on internal infrastructure and pricing agreements, but the estimate highlights just how deeply AI is embedded in daily operations.
Despite the transformation, the growing focus on token consumption raises a more nuanced question: Does more usage actually mean better work?
“There’s a difference between using AI well and just using a lot of it,” one employee told The Information. While token counts can signal depth or experimentation, they don’t necessarily reflect efficiency or skill.
According to The Information, Meta has not formally tied token usage to performance reviews.
David Adubiina
