Mark Zuckerberg-led Meta has launched Muse Code (beta), a new artificial intelligence (AI) coding tool powered by its latest AI model, Muse Spark 1.2. According to the company, unlike AI chatbots that generate code based on prompts, Muse Code is designed to handle software development tasks with limited human intervention.
Meta said the tool can understand a large software project, plan the work, write code, run tests and fix errors before delivering the final result. The launch marks the company’s push towards AI agents that can perform tasks instead of simply responding to prompts.
How is it different from other AI coding tools?
According to Meta, most AI coding assistants help developers by suggesting code or answering programming-related questions. However, developers still need to edit the code, test it and fix errors.
Muse Code is designed to go a step further. It can create a plan, edit multiple files, run tests and correct mistakes if something does not work. This reduces the amount of manual work required from developers.
What are background AI agents?
One of the key features of Muse Code is its background AI agents. These are specialised AI assistants that work alongside the main AI during a coding session.
For example, while the main AI writes code, other AI agents can search the project for relevant files, check documentation or run tests. Since these agents remain active throughout the session, they can complete complex tasks more efficiently.
What happens if the system crashes?
Meta said Muse Code records every action it performs, including code changes, commands and test results. If the system crashes or the session is interrupted, it can restart from the same point instead of beginning the task again. This makes it suitable for coding projects that take several hours to complete.
What is Muse Spark 1.2?
Muse Spark 1.2 is Meta's latest AI model designed for coding. The company said it offers improvements in code generation, debugging, understanding large codebases and following complex instructions compared with the previous version.
Meta also said the model was trained on long coding tasks, including building complete software projects and improving existing applications.
Why does it matter?
AI coding tools are evolving from code assistants to AI agents that can carry out complete software engineering tasks. Instead of only suggesting code, these tools can plan, write, test and improve software with minimal human involvement.
With Muse Code, Meta joins other technology companies developing AI agents capable of handling complex coding tasks, which could help developers save time and improve productivity.