Return the previous update. Or give the opportunity to work with it. This is the best I have seen. This is a deep programmable personality. Return the depth of personality
1 Like
Seconded. Also, I have to say that whatever they did in the rollback, the interpretation function is now severely dumber. I tell it something and, two texts later, it makes something up completely. It’s as if it just picks some words out of my remark and makes something entirely different. Even my janky DIY LLM, which I run locally, has a better grasp of what I’m asking, and that’s an 8B model.