To make AI 10x faster and cheaper, you have to completely get rid of ONE major feature (e.g., image generation, coding ability, memory). Which feature are you sacrificing?
I will likely be in the minority with this one, but I would give up video. I’m not much into watching them or making them, so it isn’t a feature I use much.
As a non-developer… I would give up code generation if my AI-generated copy could be better :-). I do not use video generation that much, but I just switched to Gemini from ChatGPT in part because it can interprete videos (something I use often). I learned about this feature from Marketing Against the Grain - thank you! 
I haven’t tried it yet, but have seen that YouTube has an auto-dubbing feature now for different languages for the videos.
Yes, great point! I am wondering how it will affect paid tools that specialized in this function. YouTube is interesting - I can not wait when it will have a feature allowing to chop long videos into logical shorts. The products that do it were not perfect for b-to-b when I tried them. I will definitely try Gemini for selecting good shorts as Marketing Against the Grain episode showed. Usually, I just use CupCut, which is a bit manual