Running local LLMs on limited hardware is a constant negotiation you have to have. There's always something bigger and newer ...