Our freedom to use, develop, and share software in general has never been more threatened than by the recent efforts to centralize and control AI. This talk will focus on what the "Local LLaMA" movement is doing to democratize AI and preserve software freedom for hackers, by developing open source projects like llamafile and llama.cpp. This talk will cover the technical challenges we've faced, making LLMs run fast on personal computers. It will also cover the organizational challenges we'll encounter facilitating respectful projects that can scale to include the same number of contributors as the Linux Kernel.