Imagine a robot that not only responds to your requests but also remembers your past conversations and reacts with genuine emotions. Researchers at the University of Geneva have brought this vision closer to reality with Nadine, a social robot powered by a large language model (LLM). Unlike typical AI assistants, Nadine goes beyond simple question-and-answer interactions. She can recognize familiar faces, recall previous encounters, and even simulate human-like emotions based on the conversation. This remarkable feat is achieved through a novel framework called SoR-ReAct, which combines the reasoning power of LLMs with affective capabilities and a memory system. SoR-ReAct allows Nadine to access external knowledge, tap into her long-term memory, and process emotional nuances, resulting in more natural and engaging interactions. When Nadine interacts with someone she's met before, she can retrieve past conversations, adding a personal touch to the experience. Her emotional responses aren't just pre-programmed reactions; they are dynamically generated based on the context of the conversation, her personality, and even her current mood. This means Nadine can express joy, sadness, anger, or other emotions in a way that feels authentic and human-like. While Nadine currently excels in one-on-one interactions, the research team is already working on expanding her abilities to handle group conversations and navigate the complexities of multi-party interactions. Nadine represents a significant leap forward in social robotics, demonstrating the potential of LLMs to create robots that are not just functional but also emotionally intelligent companions. As this technology continues to evolve, we can expect even more sophisticated social robots that blur the lines between machine and companion.