Cursor 2.0 is here... 5 things you didn't know it can do

Quick Overview

Cursor released version 2.0, introducing five major new features including the Composer Model, Agent Layout, Native Browser, Parallel agents, and the In-editor Bugbot, positioning itself as a formidable AI coding tool that surpasses many competitors in speed and intelligence benchmarks.

Key Points: Cursor released version 2.0 on October 29, 2025, featuring five major new capabilities. The new Composer Model demonstrates superior speed (around 180 tokens/second) and strong coding intelligence (Cursor-Bench score around 55) compared to frontier models like GPT-5 and Claude. The Git Worktree integration allows agents to run in parallel across separate working trees without conflicting with the main codebase. The In-editor Bugbot enables developers to point at the exact HTML element causing an issue in the native browser preview, allowing for immediate debugging. Parallel Agents allow users to run multiple LLMs (like Sonnet 4.5, GPT-5, and Composer 1) simultaneously on the same task to compare results. The Native Browser feature directly supports in-editor debugging, allowing users to inspect elements and use Chrome Dev Tools without leaving Cursor.

Context: This video serves as a review and demonstration of the major updates included in Cursor version 2.0, an AI-powered code editor that is a fork of VS Code. The creator, who previously expressed skepticism about AI coding tools being more than just a nice UI wrapper, explores five key new features intended to significantly enhance developer productivity and coding quality, comparing its new Composer model directly against competitors like Claude and GPT-5.

Detailed Analysis

Cursor 2.0 launched with five major features designed to elevate the AI coding experience beyond simple boilerplate generation. The standout feature is the Composer Model, which benchmarks show combines strong intelligence (around 55 on Cursor-Bench) with best-in-class speed (nearly 200 tokens/second), significantly outpacing models like GPT-5 and Claude in both metrics when tested on coding tasks. The video demonstrates the Agent Layout, which cleans up the chat view when running multiple agents in parallel, and the Parallel Agents feature itself, which allows side-by-side comparison of outputs from different models like Sonnet 4.5, GPT-5, and Composer on the same prompt. Another significant upgrade is the Git Worktree integration, allowing agents to operate on separate branches concurrently without interfering with the main workspace. Furthermore, Cursor now includes an In-editor Bugbot paired with a Native Browser view, enabling users to click on visual elements in the live preview, inspect the corresponding HTML/CSS in the Dev Tools overlay, and instruct the AI to make direct fixes, proving to be a substantial productivity boost compared to traditional external debugging loops. The creator also contrasts the quality of code generation for neobrutalism buttons and Apple's liquid glass design across Claude, GPT-5, and Composer, concluding that Composer showed the most potential despite GPT-5 sometimes yielding slightly better results in specific visual tests.