I tested every local LLM tweak people recommend, and only these ones actually mattered
…Then I actually started messing with the settings myself, and it changed how I think about using a local model entirely. The defaults aren’t bad, but they’re also not tuned…