>>105935552
it's not the card (model prompts), but the LLM running it that makes or breaks her as a chatbot