Making AI Thought Process Collapsible

Feature Requests
1

Current Pain Points

When using DeepSeek-R1 and Gemini-flash-thinking for code generation/Q&A:

  1. AI’s detailed step-by-step thinking occupies significant screen space
  2. Verbose intermediate reasoning steps make it harder to quickly locate the final output
  3. Lengthy thought processes reduce reading efficiency during iterative debugging

Proposed Solution

Add the following to code generation/Q&A result displays:

[+] Thought Process (Click to expand)
  1. Default collapsed state for full AI reasoning path
  2. Interaction features:
    • Click to expand/collapse individual thought processes
    • Global “Expand All/Collapse All” toggle
  3. New settings options:
    • Always show full thought process
    • Default collapse secondary reasoning steps
    • Completely hide thought processes

Typical Use Cases

  • Quick code review scenarios (e.g., generated functions/components)
  • Rapid iteration during debugging sessions
  • Sharing clean code snippets with teammates
  • Small-screen device usage

Bonus Suggestions

  1. Show key step summaries in collapsed view
  2. Implement keyboard shortcut (e.g., Cmd+/) for display mode switching

Hey Cursor Team, this would be a game-changer for focused coding sessions! Thanks for your continuous improvements!

2

Hey, as we begin to integrate further with thought models like DeepSeek R1, we are hoping to have some improvements around the UI here, to ensure you see only what you want, without the Composer being filled up by the thinking step of the model!

3

Now the thought process is not visible anymore?