Meta Reinforces Its AI Bet: Zuckerberg Uses Claude to Write Code, Employees Kick Off a Token Consumption Battle to Hit Their KPIs

Meta is doubling down on AI—Zuckerberg is personally writing code with Claude Code. To hit KPIs, an internal Token-consumption chaos has emerged, wasting resources. Taking the lesson from the metaverse loss of $80 billion, Meta is actively acquiring startups, hoping to turn the technology into tangible value.

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

Meta, a tech giant with social platforms such as Facebook, Instagram, and Threads, has recently been putting the company’s resources entirely into the generative AI space.

Meta CEO Mark Zuckerberg has started using an AI code-writing tool, Claude Code, to write code himself—breaking his record of not participating in direct development for many years.

Foreign media have also reported that a Token-consuming contest has recently broken out inside Meta. Many engineers raise their personal performance metrics (KPIs) by consuming large amounts of Tokens.

AI code-writing is all the rage, and founders are returning to the development frontline

In March 2026, Zuckerberg submitted 3 code diffs to Meta’s single repository—his first substantive code contribution in 20 years.

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

His actions reflect how AI code-writing tools are attracting corporate founders to re-enter systems development. Y Combinator CEO Garry Tan has also returned to code writing after 15 years and has open-sourced a system that combines Claude Code.

According to internal documents from Meta that leaked in March 2026, the company has set aggressive goals and plans for 65% of engineers to use AI to write 75% or more 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’s internal Token-consumption competition turns KPIs into performance

To promote the adoption of generative AI applications, a phenomenon has emerged inside Meta that links Token usage to productivity. Tokens are the smallest units of text processed by large language models; in Chinese, they are often referred to as “symbol tokens” or “word tokens.”

《The Information》 reported that inside Meta there is a leaderboard called Claudeonomics, 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, with the top user’s average consumption reaching 281 billion tokens.

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

A 《Forbes》 report notes that Meta’s Chief Technology Officer Andrew Bosworth has mentioned that a top engineer consumes a quantity of Tokens roughly equal to his annual salary; Nvidia CEO Jensen Huang has also said that if an engineer earning $500,000 cannot consume $250,000 worth of Tokens, he would be concerned.

However, this KPI system that drives internal Token consumption to spiral also brings downsides. To boost performance numbers, some Meta employees leave AI agent programs running idly for hours, wasting computing resources.

Moreover, treating employees’ Token consumption directly as a productivity metric makes the consumption behavior into something performed for appearances, leaving performance reviews facing the challenge of lacking support from substantive business outcomes.

Learning from metaverse failure, Meta’s next challenge in the AI race

Before making a major push into AI, Meta’s re-pledged metaverse bet ended in failure. The company previously poured about $80 billion into building the virtual world Horizon Worlds and VR/MR devices, and even changed the company name to “Meta,” but ultimately still failed to reach the user scale expected by the market.

Lily Liu, president of the Solana Foundation, also expressed a pessimistic view in the comment section of a social platform discussion about blockchain gaming and metaverse development—criticizing the virtual economy models of the past for lacking substantive content to support them.

Image source: Meta; Meta’s metaverse platform Horizon Worlds. In the initial version, Zuckerberg’s virtual likeness shown

Now that Meta is shifting its focus to AI, it is actively planning its market layout. In addition to rolling out 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 acquired Moltbook, a proxy community hailed as the “AI version of Reddit.” Moltbook’s founders Matt Schlicht and Ben Parr will join Meta’s team.

• **Related report:**The “Lobster community” was bought! Rumor: Meta acquires Moltbook; the founder successfully switches from media to AI

Outside observers are still watching to see whether Meta can avoid repeating the metaverse’s pattern of overinvestment without substantive applications—turning the current internal Token-consumption frenzy and acquisition deals targeting startups such as Moltbook into actual products with commercial value, and thereby securing a foothold in the fiercely competitive generative AI market.