Core Insight
Cursor's success came from recognizing that AI-assisted coding is fundamentally a UX problem, not just a model problem. By rebuilding the editor from scratch around AI interactions rather than bolting AI onto an existing IDE, they created workflows that feel native rather than forced.
My Analysis
The key product decisions that stood out:
- Tab completion as the killer feature — not chat, not code generation, but inline suggestions that feel like a mind-reading autocomplete. This reduced the friction of AI assistance to near zero.
- Fork of VS Code — controversial but brilliant. They got the entire extension ecosystem for free while having full control over the core editing experience.
- Composer mode — multi-file editing that understands project context. This is where the real value emerges vs GitHub Copilot's single-file suggestions.
The lesson for AI products: start with the interaction model, not the model itself.