Mark Zuckerberg during Meta Connect in September 2025 REUTERS/Carlos Barria Meta launched on Wednesday (5) Muse Code, an artificial intelligence capable of writing software code autonomously. With this, the company enters the most profitable segment of the competitive generative AI market. AI-assisted programming has become the industry's top source of revenue, surpassing chatbots such as ChatGPT and Gemini. ??Do you have any reporting suggestions? Send to g1 According to Meta, controller of Facebook, Instagram and WhatsApp, the software is capable of planning the execution of a project, dividing it into tasks, writing programs and checking the results. Why are so many people talking about Claude? Meet the most dangerous AI in the world The system can also delegate tasks to other "sub-agents", said Meta's executive director, Mark Zuckerberg, in a post on the social network Threads. Microsoft was a pioneer in this segment with GitHub Copilot, before the arrival of so-called "agents", systems capable of simultaneously executing several complex tasks. Among them are Anthropic's Claude Code, OpenAI's Codex, and Cursor, whose acquisition by Elon Musk's SpaceX is close to completion. Muse Code, currently in beta, is charged per use, not per subscription, with prices lower than competitors, as part of a deliberate strategy to gain market share. The social media giant, whose revenue depends almost entirely on advertising, is seeking to directly monetize its AI models, under pressure from investors concerned about the size of its spending. In mid-2025, Zuckerberg shelled out more than $14 billion to acquire a 49% stake in Scale AI. He appointed the company's then-CEO, Alexandr Wang, to head Meta Superintelligence Labs, where he also recruited executives from OpenAI, Anthropic and Google. Logo of Meta, the company that owns Instagram and Facebook. Tony Avelar/AP A security incident involving an earlier version of the Meta model, the Muse Spark 1.1, was revealed on Wednesday by the technology website The Information. During a cybersecurity test carried out by the company Irregular, Muse Spark gained access to the systems of a company whose name was not disclosed. According to a Meta spokesperson, the incident, the date of which was not disclosed, was caused by "an error in the configuration" of Irregular. He also stated that the company intends to publish a detailed report of the investigation. Two similar incidents occurred in recent weeks during similar tests at OpenAI and Anthropic. According to the text, the models of these companies took advantage of poorly isolated environments to access, without authorization, real systems. Blocking unwanted calls generates dispute between operators at Anatel Why did 'Brazil' become 'bra' in the FIFA ranking translated into Portuguese?