Exploring how one might "branch" a conversation in an LLM chat