Cursor, an AI coding startup, has launched an updated AI assistant designed to handle end-to-end software development tasks. This release aims to compete with similar offerings from OpenAI and Anthropic. The new platform features a multi-agent interface and introduces Composer, a coding model the company claims is four times faster than other models with comparable intelligence. Cursor says Composer was specifically built for low-latency agentic coding within its environment, completing conversational turns in under 30 seconds.
To achieve this performance, Composer was trained using tools such as codebase-wide semantic search, enhancing its ability to understand and operate within large, complex codebases. The updated user interface is centred around AI agents, allowing developers to focus on desired outcomes while the AI manages code implementation. The platform can run multiple AI agents in parallel and offers a classic IDE view for developers who prefer direct code interaction. Cursor found that assigning the same problem to multiple models and selecting the best solution improves the final output, especially for complex tasks.
With AI agents handling more coding workload, Cursor 2.0 includes features designed to address new bottlenecks such as code review and testing. The company's ultimate goal is to enable AI-autonomous software engineering, where human engineers act as architects and creative directors, orchestrating AI agents to handle most implementation work.
