Meta Internal Document Leaks AI Coding Hard Metrics: 65% of Engineers Must Use AI to Write Over 75% of Code

According to monitoring by 1M AI News, an internal document from Meta shows that the company is setting specific goals for the use of AI tools across its business departments, which is the latest initiative by CEO Mark Zuckerberg to push Meta to become an “AI-native” company.

The most aggressive metrics in the document come from the Creation team responsible for core creative experiences: by the first half of 2026, 65% of engineers must use AI to complete over 75% of their submitted code. The Scalable ML team has set a target for February 2026 of having 50% to 80% of code completed with AI assistance, but a senior engineering manager noted that “we are not tracking this through metrics.”

At the company level, the document outlines targets for the fourth quarter of 2025 covering core products like Messenger, WhatsApp, and Facebook: 80% of mid- to senior-level engineers must adopt AI tools like DevMate, Metamate, and Google Gemini (focusing on “tool adoption” rather than the proportion of AI-generated code), and 55% of code changes must be completed with “agent assistance.” It is currently unclear whether these targets are tied to performance evaluations.

CTO Andrew Bosworth announced this week that he will personally oversee Meta’s “AI for Work” project, promoting the comprehensive adoption of internal AI tools. Some Reality Labs employees have been reassigned new titles such as “AI Builder,” “AI Pod Lead,” and “AI Org Lead,” reflecting the company’s shift towards smaller teams and a flatter organizational structure. A Meta spokesperson stated that the performance system focuses on the actual impact brought by AI tools, rather than just usage rates.