root since 2005 frontend since 2016 now: GPUs, containers and LLM inference in a home lab I break on purpose real benchmarks, raw terminal output