As you may know, many Cursor users rely on tools like ChatGPT to help the agent better understand prompts or provide clearer instructions.
For example, I often use ChatGPT as an expert code reviewer and sometimes as a programmer, while I act more like a product managers seating between the two products.
I frequently use the “Copy Message” button in Cursor to feed Cursor’s work into ChatGPT for code and instruction reviews.
The issue I encounter is that the “Copy Message” button is only available after accepting changes. However, there are times when I’d like to copy the proposed changes from Cursor into ChatGPT for review before accepting them. This would be especially helpful for enhancing and refining the code further.
While I’m not a developer, I’m technical enough to understand web apps, and I want to share how Cursor is truly transforming my workflow.
It would be fantastic to have the ability to copy the proposed message before accepting changes.
While I understand what you’re saying and its usefulness, I think it’s easier to debug changes within Cursor’s chat window that has access to the full codebase (and .cursorrules file) for more context than in ChatGPT.
The downside is that ChatGPT allows you to review changes before accepting them, which I believe is the crux of why you’re implementing this external checl.
I think a better solution is to accept the changes within Composer. If the changes aren’t working as expected, you can:
Debug it within the chat window
Debug it within the Composer window
Revert to a checkpoint
This way you’re able to:
Better see how it integrates within your codebase
Ask follow-up questions to a model with more context
Revert changes if necessary by reverting to a checkpoint
To synthesise my response: What advantage does this external ChatGPT check offer as opposed to just accepting the changes and then reverting to a previous checkpoint if it doesn’t go as planned?