Feature request for product/service
AI Models
Describe the request
I’ve had the experience that Cursor - or any coding assistant for that matter - is unable to properly analyse certain tasks, or rather deeply enough. When the request is something like “please create an algorithm that does this and that” the response is often superficial. E.g. if the proper solution would require deeper mathematics or complex, heuristic or gradient descent algorithms (not neural networks, just simple optimization problems) then it is unable to suggest complex enough solutions. In these case, if I take the problem to an AI with a deep research ability, like Gemini, they are able to find me the correct algorithms, and explain to me how they are implemented. If then I take this response back to Cursor, then it is also able to implement them without any issues. But not by itself.
It would be interesting to see if Cursor could do such deep research too, without asking for the input from a third party AI.
Of course this carries the danger that users would try to solve simple problems with it, even when it is not needed, increasing the token usage by a lot. So somehow the usage of this feature should be automated, or users should be discouraged somehow from using it all the time.
Operating System (if it applies)
Windows 10/11
MacOS
Linux