Meta re-doubles down on AI: Zuckerberg uses Claude to write code, while employees launch a token consumption battle to hit KPIs

Meta is doubling down on AI, with Zuckerberg personally using Claude Code to write code. To hit KPIs, a token-consumption chaos has apparently broken out internally, wasting resources. Learning from the lesson of the Metaverse’s $80 billion loss, Meta is actively acquiring startups, hoping to turn technology into real value.

Meta doubles down on AI: Zuckerberg writes code with Claude Code

Meta, a tech giant that owns social platforms like Facebook, Instagram, and Threads, is recently putting the company’s resources fully into the generative AI field.

Meta CEO Mark Zuckerberg has started personally using an AI code-writing tool, Claude Code, to write code, breaking a record of many years in which he did not directly participate in development.

According to reports from overseas media, a token-consuming competition has recently also emerged inside Meta, where many engineers drive up their personal performance indicator (KPI) by consuming large amounts of tokens.

AI code-writing is all the rage, and the founder returns to the development front line

In March 2026, Zuckerberg submitted three code diffs to Meta’s single repository—his first substantive code contribution in 20 years.

The terminal-based coding assistant Claude Code CLI he uses is developed by Anthropic, and one of his submissions received approval from more than 200 engineers.

His actions reflect that AI code-writing tools are attracting company founders to get back into system development. Y Combinator CEO Garry Tan also returned to coding work after 15 years, and open-sourced a system that combines Claude Code.

According to Meta internal documents leaked in March 2026, the company has set ambitious goals, planning for 65% of engineers to use AI to write more than 75% of their code by mid-2026.

Image source: flickr, photo by Niall Kennedy Meta founder Zuckerberg speaking at the Facebook F8 developer conference in September 2011

Meta sparks a token-consumption contest internally, and KPIs become a show

To promote generative AI applications, a phenomenon has emerged inside Meta that links token usage to productivity. Tokens are the smallest units that large language models process text with; in Chinese, they are often called “symbol units” or “tokens.”

《The Information》 reports that inside Meta, there is a leaderboard called Claudeonomics that tracks AI token consumption across more than 85k employees. The data shows that within just 30 days, employees consumed as many as 60 trillion tokens, with the top user averaging 281 billion tokens consumed.

The leaderboard also sets titles such as Token Legend to encourage employees to integrate AI tools into their day-to-day work.

《Forbes》 reports that Meta’s Chief Technology Officer Andrew Bosworth previously mentioned that a top engineer consumed a token amount equal to their annual salary; Nvidia CEO Jensen Huang has also said he would be concerned if an engineer making $500k in salary failed to consume tokens worth $250k.

However, this kind of KPI regime that drives token consumption “competition for its own sake” also brings drawbacks. Some Meta employees, to boost performance figures, leave AI agent programs running for hours while idle, causing waste of computing resources.

Moreover, treating employees’ token consumption amounts directly as a productivity metric turns consumption behavior into performance art, creating challenges for performance evaluations that lack support from tangible business outcomes.

Learning from the Metaverse’s failure experience, Meta’s next challenge on the AI track

Before making a big push into AI, Meta’s re-bet on the Metaverse ended in failure. The company once invested about $8 billion to build the virtual world Horizon Worlds and VR/MR equipment, even changing the company name to “Meta,” but ultimately still failed to reach the market-expected scale of users.

Lily Liu, president of the Solana Foundation, in the comment section on a social platform, also shared pessimistic views on the virtual economy model in the development of blockchain games and the metaverse—citing past lack of substantive content backing.

Image source: Meta Meta’s Metaverse platform Horizon Worlds In the earliest version, shown is Zuckerberg’s virtual likeness

Now that Meta is shifting its focus to AI, it is actively laying out its market strategy. In addition to launching its own large language model LLaMA, it is also gradually advancing an AI model initiative called “Avocado.”

A recent report from 《Axios》 also revealed that Meta has acquired Moltbook, a so-called “AI version of Reddit” proxy community. Moltbook founders Matt Schlicht and Ben Parr will join Meta’s team.

• **Related report:**The Lobster community was bought out! Report: Meta acquires Moltbook, and the founders successfully transition from media to AI

The outside world is also watching whether Meta can avoid repeating the Metaverse’s mistake of overinvesting without real applications—turning the current internal token-consumption frenzy and acquisition deals like those involving Moltbook into actual products with commercial value, so it can stand above in the fiercely competitive generative AI market.