I still don’t see much of improvement in Gemini 3

I have been using Ai studio for a while mostly for helping in writing ML code, there were times where 2.5 pro made mistakes.

Now I see some of the mistakes similarly made by 3 pro. For eg. Forgetting about importing a very simple library or just following instructions I said.
Even when I give API refrance for new library it ends up using old methods, there are times where it just implement a custom function even though it's can be imported from a library.
And it stills feels so impatient and trying to take shortcut and tbh it's not just Gemini I faced this problem with other AI also.

ALSO I hate this constant manipulation gemini does instead of being unbias even after giving so many instructions.

It will be very helpful for any insight as maybe I can improve my prompt or any custom system instructions you guys use.

It is better than 2.5 pro but these are some of the short comming I am facing.

Leave a Reply