Right I also thought of running a local paired down deepseek model or pregenerating. Honestly for a silly thing like this I want to keep the footprint as low as possible. And currently I would have to host a local model on a literal potato in the form of an rpi3b+ running omv6 and docker and let me tell you that is even more sluggish than you’d imagine. I have a few bucks of api left though and could craft a roleplay persona with rules for all the cat text and also fake facts. Had gemini flash make a little animated ascii cat script in python last night and it could take some time but I think I could finesse a few loops that way though I wonder if I shouldn’t do the ascii by hand just to save the headaches of prompting to align things correctly and such. Also got to figure out how to make it live in the terminal without displacing the normal text and allow it to seem to interact with what is there. I’ll keep picking at it. It’s kindof the side thing I’m working on when I run out of daily gemini cli quota. We will see how it goes