Cursor Outpaced Copilot By Rebuilding IDE
Zach Lloyd, CEO of Warp, on the 3 phases of AI coding
Cursor won the first wave of AI coding by making code completion feel native to the act of editing, not like a helper bolted onto the side. Copilot proved demand for autocomplete, but it lived mainly as a plugin inside existing editors. Cursor rebuilt the editor around low latency edits, larger context windows, and chat driven changes, which made the product feel closer to typing with an extra brain than accepting occasional suggestions.
-
Copilot launched in June 2021 as an AI pair programmer inside Visual Studio Code. That gave it huge distribution through GitHub, but the product shape was still an add on to an existing IDE workflow, centered on inline suggestions and later chat.
-
Cursor started from the opposite assumption. Instead of inserting AI into someone else's editor, it forked the IDE and optimized the whole loop around prediction, edit application, and agent style workflows. That let it move from autocomplete into Composer and broader task execution without changing the core product.
-
The market response shows how strong that product advantage was. Cursor grew from about $65M ARR in November 2024 to $100M by early 2025, then to $200M in March 2025 and $500M in May 2025. That speed turned a feature battle into a platform battle across IDEs, terminals, and coding agents.
The next step is convergence. The winning tools will keep the fast, almost telepathic edit experience that made Cursor break out, then layer on agents that can plan, modify, run, and debug across many files. That shifts competition away from simple autocomplete quality and toward owning the full developer workflow, from prompt to diff to terminal execution.