I'm just training LLMs with datasets retrieved over multiple peer-to-peer protocols.
Your LLMs must be great at regurgitating full linux distro images :)