Can someone with no programming background build production-grade software using an AI code editor? That's the question product manager Pavel Manovich set out to answer, documenting his experience building a full-scale application using Cursor.
The experiment sits at the center of a growing debate in the developer tools space. Cursor's agent mode - where you describe what you want in plain English and the AI writes, edits, and refactors code across multiple files - has gotten good enough that non-technical founders and product people are attempting projects that would have required a development team 18 months ago.
The results are more nuanced than the "anyone can code now" narrative suggests. AI coding tools handle the straightforward parts well: scaffolding a project, wiring up standard UI components, connecting to APIs with good documentation. Where they consistently struggle is with complex state management, debugging subtle issues across multiple services, and making architectural decisions that won't create problems three months later.
This matches what we've seen across the AI coding tool market. Cursor, Claude Code, GitHub Copilot, and Bolt have all made it faster to get from zero to a working prototype. The gap between a working prototype and maintainable production software is where the "vibe coding" approach runs into trouble.
For non-developers considering this path: the tool has genuine value for building internal tools, MVPs, and automations. Building something you plan to scale to thousands of users still requires someone who understands software architecture, even if the AI writes most of the actual code.