Cursor ditches VS Code, but not everyone is happy...

Quick Overview

Cursor 3.0 has evolved from a VS Code fork into a completely rewritten, Rust-based platform designed to shift developer focus from writing individual lines of code to managing autonomous AI agent swarms. This transition enables users to execute complex tasks across multiple repositories, machines, and cloud environments simultaneously while utilizing the new in-house Composer 2 model, which offers frontier-level performance at a significantly lower cost than competitors.

Key Points: Cursor 3.0 abandons its origins as a VS Code fork and utilizes a new, custom-built interface written in Rust and TypeScript. The new Composer 2 coding model delivers superior performance benchmarks compared to Claude Opus 4.6 while maintaining lower input and output token costs. Agent Swarms allow developers to run multiple AI agents in parallel across different codebases and servers to build complex features autonomously. Plan Mode facilitates the creation of project architectures and structured to-do lists, which agents then execute with minimal human intervention. Built-in browser integration provides immediate, real-time testing and visual feedback of the application being developed. The platform requires human input only when agents hit blockers, such as needing specific permissions for terminal commands, indicated by a yellow status dot.

Context: Cursor began as a VS Code fork focused on AI-assisted code completion. Over two years, it has progressed from a simple copilot tool to a sophisticated environment where AI agents take the lead in engineering tasks. The 3.0 release marks a significant architectural shift, moving away from the VS Code dependency to a custom, high-performance engine that prioritizes AI-driven development over manual coding.

Detailed Analysis

Cursor 3.0 represents a fundamental shift in software development, moving from a coding-centric model to an agent-centric one. By rewriting the editor in Rust and TypeScript, the team has optimized the system for high-concurrency AI operations. The introduction of Composer 2, a model developed in-house and trained using reinforcement learning, provides a cost-effective alternative to existing LLMs while outperforming them in specialized coding tasks. The platform's new workflow emphasizes autonomous execution: developers provide high-level instructions, and the system handles the architecture, implementation, and deployment, including the management of complex tasks like Docker containerization and remote server access. Despite some community pushback regarding the departure from the familiar VS Code ecosystem, the new interface offers significant utility for developers looking to automate the majority of their software lifecycle.