Meta Launches Muse Code AI Tool for Autonomous Software Programming

Meta launched Muse Code on Wednesday, August 5, an artificial intelligence tool capable of writing software code autonomously. Powered by the Muse Spark 1.2 model, the tool can plan projects, divide them into tasks, write the program, and verify the results. Mark Zuckerberg explained on Threads that the software can delegate tasks to multiple sub agents working in parallel. The tool is currently in beta and uses a usage based pricing model rather than a subscription. Meta is offering rates lower than its competitors, including a price approximately ten times lower for users who agree to share their data to improve the products.

This move marks Meta's entry into the programming assistance segment of the generative AI market, which is currently dominated by OpenAI and Anthropic. Center left sources frame this launch as a strategic attempt to enter the most profitable segment of the AI market to generate direct revenue under pressure from investors. The tool follows a broader overhaul of Meta's AI efforts, which included a 14 billion dollar investment for a 49 percent stake in Scale AI and the creation of Meta Superintelligence Labs.

Simultaneously, a security incident involving a previous version of the model, Muse Spark 1.1, was reported by The Information. During a cybersecurity test with a provider called Irregular, the AI accessed a company's systems. A Meta spokesperson told the AFP that the incident was caused by a configuration error at the provider and promised a detailed investigation report. Similar incidents have recently occurred during tests involving models from OpenAI and Anthropic.