Summary
Fireship covers Cursor 3.0, the most significant release in the AI coding editor’s history — and the one that ends its identity as a VS Code fork. The new version is a ground-up rewrite in Rust and TypeScript, repositioning Cursor from a code editor into an agent orchestration platform. Users can now run multiple AI agents in parallel across different repositories, remote SSH servers, and cloud environments from a single unified window, with color-coded status indicators and a built-in browser for live app preview.
The release was accompanied by Composer 2, Cursor’s in-house model, which sparked controversy when users discovered its model ID in metadata — revealing it was based on Moonshot AI’s Kimi K2, not an independent model as initially implied. The situation added a layer of irony: Kimi K2 itself has been accused of training on Claude outputs, occasionally identifying itself as Claude. Cursor acknowledged the lack of transparency, apologized, and published a technical report on their reinforcement learning fine-tuning approach.
Fireship demonstrates the new interface through a multi-agent build project, highlighting plan mode, parallel agent management, design mode for visual iteration, and the built-in browser. Community criticism comparing the new direction to OpenAI Codex is acknowledged. For developers tracking the competitive landscape of AI coding tools, this video captures a pivotal product and reputational moment for one of the category’s leading players.
📺 Source: Fireship · Published April 06, 2026
🏷️ Format: News Analysis
