I wonder if 1 prompt could be processed by 2 LLM and let us compare the result,

there already some tools out there could do so, the user have to provide his/her own APIs.

imagine if cursor do the same (yes it used 2-3x more the credit),
but each time, you can SWITCH among LLMs, that will be wonderful.

As i found sometimes 1 LLM work better for some e.g. c3.7 better for
graphics/physic while G2.5Pro for other things.

oops… seems someone discussed about multi agents before,
similar.