Comparing Copilot Edits and Cursor
| Feature / Area | GitHub Copilot Edits (Observations) | Cursor (Observations) |
|---|---|---|
| File Context Selection | Limited to selecting only 10 files at a time. | User experience for including files as context is much better. |
| Applying Changes (VS Code) | Requires manually saving each file after changes. | (Implied smoother workflow / less manual saving compared to GC Edits in VS Code). |
| Agent Mode Power | Less emphasis or specific powerful "Agent mode" noted. | Agent mode looks more powerful. |
| Agent Capabilities | Basic editing/generation. | Agent can list/read files, run commands, check linters, etc. - feels SOTA. |
| Agent Limitation | Not specifically mentioned. | Agent struggles with commands needing sudo access. |
| Suggestions/Completion | Standard Copilot suggestions. | Tab suggestion and completion functionality is "mind blowing". |
| Edit/Composer Behavior | Tends to give code changes even when just asking a question. | Tends to give code changes even when just asking a question. |
General Tips for Using AI in Coding:
- Using statically typed languages works better with AI; linters catch issues earlier.
- For dynamically typed languages, add comments specifying input/output types.
- Always plan your code changes before using the AI; this improves prompt quality.
- AI models often struggle with adding the correct library version in dependencies.
- Keep file sizes small for better AI context understanding.
Note: Things observations are from March 2nd 2025
Top comments (2)
This is no longer true.
Care to elaborate?
Definitely the most annoying "feature".
Agent mode is much more powerful because it can figure out things on it's own, run commands like grep, find etc., read linter errors and run commands.
I will rewrite this article with latest analysis.