In a post-apocalyptic setting, a small, off-grid language model running on something like a Raspberry Pi is presented as a survival tool, with one startup charging 7x the MSRP for an Nvidia Jetson Orin Nano with a battery enclosure. However, the article argues this concept is better left to roleplaying a zombie apocalypse, as a local LLM could be a liability in a real survival situation. While these models can retain information, such as the entire English text of Wikipedia (about 24.7 GB compressed), and support vision, translation, and speech capabilities, they still hallucinate and fabricate information. LLMs are trained on tens of trillions of words, but their data retrieval is imperfect, functioning like auto-complete. Modern chatbots from OpenAI, Anthropic, or Google attempt to mitigate hallucinations by pulling in live web data, but there is no guarantee the retrieved information is accurate. Medical professionals at Duke University School of Medicine warn that models cannot understand patients' underlying meaning and often have sycophantic tendencies. According to OpenAI, more than 230 million people globally query ChatGPT with health and wellness related questions each week.