Meta Re-Puts Its AI Bets: Zuckerberg Uses Claude to Write Code, 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 meet KPIs, an internal token-consumption frenzy has emerged, and resource waste has become rampant. Taking lessons from the metaverse’s $80 billion loss, Meta is actively acquiring startups, hoping to turn the technology into real value.

Meta is doubling down on AI: Zuckerberg writes code using Claude Code

Meta, a tech giant that owns social media platforms such as Facebook, Instagram, and Threads, is currently committing all of the company’s resources to the generative AI space.

Meta CEO Mark Zuckerberg (Mark Zuckerberg) has begun personally using the AI coding tool Claude Code to write code, breaking a years-long record in which he had not directly participated in development.

According to reports from overseas media, Meta’s internal teams have also recently sparked a competition focused on consuming tokens, with many engineers using large amounts of tokens to boost their personal performance indicator (KPI) numbers.

AI coding is all the rage—founders return to the front line of development

In March 2026, Zuckerberg submitted 3 code diffs to Meta’s single repository—the first time in 20 years that he has made a substantive contribution to code.

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

His actions reflect that AI coding tools are attracting enterprise founders to re-engage in system development. Y Combinator CEO Garry Tan has also returned to coding after 15 years, and has open-sourced a system that integrates Claude Code.

According to internal documents that leaked from Meta in March 2026, the company has set aggressive targets and plans that by mid-2026, 65% of its engineers will use AI to write 75% or more of their code.

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

Meta holds an internal token-consumption contest—KPIs become a show

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

The Information reports that inside Meta, a leaderboard called Claudeonomics has appeared, tracking the AI token consumption of more than 85,000 employees. The data shows that employees consumed as many as 60 trillion tokens in just 30 days, and the user at the top averaged 281 billion tokens consumed.

The leaderboard also established 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 has mentioned that a top engineer consumed a token amount equal to their annual salary; Nvidia CEO Jensen Huang has also said that if an engineer earning $500,000 fails to consume tokens worth $250,000, he would feel concerned.

However, this KPI system that drives an internal race to consume tokens also brings downsides. For the sake of boosting performance numbers, some Meta employees let AI agent programs run idly for hours, causing wasted computing resources.

Moreover, treating employees’ token consumption directly as a productivity metric has turned consumption behavior into a performance, making performance evaluations face the challenge of lacking support from real business outcomes.

Learning from the metaverse’s failure experience—Meta’s next challenge on the AI track

Before investing heavily in AI, Meta’s metaverse bet ended in failure. The company previously spent about $8 billion to build the virtual world Horizon Worlds and VR/MR equipment, and even changed the company name to “Meta,” but ultimately still failed to achieve the user scale expected by the market.

When Solana Foundation President Lily Liu commented in the comment section on a social platform about the development of blockchain games and the metaverse, she also expressed a pessimistic view of the virtual economy model in the past that lacked substantive content to support it.

Image source: Meta Meta’s metaverse platform Horizon Worlds. In the original version, Zuckerberg’s virtual avatar is shown.

Now that Meta has shifted its focus to AI, it is actively arranging its market strategy. In addition to launching its own large language model LLaMA, it is also gradually advancing an AI model project called “Avocado.”

A recent report by Axios also revealed that Meta acquired Moltbook, a proxy-agent community praised as the “AI version of Reddit.” Moltbook founder Matt Schlicht and Ben Parr will join Meta’s team.

• **Related report:**The lobster community was bought! It is rumored that Meta acquired Moltbook—founders successfully transitioned from media to AI

Outside observers are also watching whether Meta can avoid repeating the metaverse’s pattern of over-investing without real applications, and instead turn its current internal token-consumption frenzy and acquisition deals involving startups such as Moltbook into actual products with commercial value, so it can secure a foothold in the fiercely competitive generative AI market.