Meta is steadily decreasing its dependence on costly AI programming tools sourced from Anthropic and OpenAI. However, it grapples with the issue of preventing employees from becoming excessively reliant on these external resources, a situation that could impede the progress of its in-house tool development. To address this, Meta has prohibited engineers from utilizing Claude Code and Codex tools and has halted certain projects that necessitate the use of these models. This move aims to prevent the integration of third-party model content into its proprietary training datasets. Additionally, Meta has formed an Applied AI Engineering team dedicated to refining its self-developed code assistant, MetaCode, as well as constructing datasets and test question repositories. Although the use of external tools is permitted in select scenarios, engineers are required to craft their own test question repositories independently and refrain from directly implementing AI-generated concepts.