Meta pushes internal engineers to use MetaCode, catching up with OpenAI and Anthropic

Odaily News: Meta is pushing thousands of its internal engineers to use its self-developed AI coding assistant, MetaCode, to enhance the code generation capabilities of its AI models and accelerate its race to catch up with OpenAI and Anthropic in the AI coding space.

According to reports, Maher Saba, Meta's Vice President of Applied AI Engineering, asked company engineers in an internal memo last month to use MetaCode to complete at least one code "diff" (code change) per week. These code modifications, once reviewed, will be integrated to improve the AI models.

Meta hopes that by deploying its AI coding tools at scale in real-world software development environments, it can accumulate more engineering practice data and improve model performance in code generation, debugging, and software development tasks.

AI coding has become a key battleground in large model competition. Tools such as Anthropic's Claude Code and OpenAI's Codex are gaining traction among developers, and Meta aims to narrow the gap between its AI coding capabilities and those of industry leaders through large-scale internal usage.

Meta has continued to ramp up its investment in artificial intelligence in recent years, building out an enterprise-grade AI ecosystem through self-developed large models, AI infrastructure, and development tools. The push to broadly adopt MetaCode internally is also seen as a significant step for the company to enhance the quality of model training data and practical application capabilities. (The Information)