Cursor backlash hits agent-first UX

Cursor is running into the hardest problem in AI devtools: once model quality converges, bad workflow design hurts more than model gains help.

• –Cursor’s own homepage now pushes agents, cloud agents, code review, CLI, and multi-surface workflows, showing how far the product has moved beyond simple in-editor assistance
• –Theo’s complaint is fundamentally about trust in the daily loop: if mode switching and UI behavior feel inconsistent, developers stop treating the tool like an extension of their editor
• –Community discussion around AI coding tools shows the same divide: some developers love agentic workflows, while others think newer defaults feel slower, noisier, and less predictable
• –That matters more for Cursor than for model vendors, because its edge is product UX and integration quality, not exclusive access to frontier models
• –The bigger takeaway is category-wide: AI IDEs can’t just get smarter, they have to stay legible under pressure for people shipping real code all day