AI's critique each other

Feature Requests
1

Would love to have Deepseek, ChatGPT, and Claude argue and critique each other for generated code and human pick the best if the AI’s don’t agree.

2

I’ve done a bunch of tests with o1 and r1 picking each other’s output apart and it I really like the outcome’s. An extremely useful tactic for sure.

3

I take it Cursor does not allow me to swap between models?

4

Hey, while you can’t necessarily get them to compete directly, you are able to switch models within the same conversation.

Hypothetically, you could make a situational test, where you first get Model A to respond, switch to Model B and say “The attached conversation history is from a chat between me, the user, and Model A. Evaluate it’s answer, and correct any errors”.

Might not be exactly what you are hoping for, but food for thought on what is possible inside Cursor.

5

I love when o1 and r1 battle it out.

These kind of prompts have given me some interesting results:

" I want you to do the following with the other’s proposal:

  1. Tell me what you strongly disagree with

  2. Tell me what you disagree with

3, tell me what you are neutral about

  1. Tell me what you agree with

5 tell me what you strongly agree with

  1. Tell me what you think they missed

  2. Tell me how you can make the plan better

I will share this back with them and we will go several rounds back and forth until we have good consensus on a plan that will be optimized"

—

“I want you to rewrite your entire proposal but incorporate any new thinking now that you have seen their feedback. Be as detailed and technical as possible will all the fully working code needed.”