Meta enters AI coding race with autonomous agent Muse Code

The agent operates on Meta’s existing Muse Spark model and simplifies the development cycle by automating the planning, coding, and validation phases. According to CEO Mark Zuckerberg, the system addresses the challenge of large-scale codebases by launching isolated sub-agents that work concurrently in separate worktrees, preventing conflicts between simultaneous tasks. During internal testing, the tool successfully implemented six features for a game engine without interfering with the primary working copy.

This release signals a strategic shift for Meta as it seeks to monetize its AI infrastructure beyond advertising. Alexandr Wang, head of Meta Superintelligence Labs, noted that the company intends to compete directly with platforms like Claude Code by offering a more affordable alternative for enterprise workflows. Following their earlier entry into customer service automation, this move marks a significant expansion of Meta’s enterprise-facing AI portfolio.