I've been seeing LLMs act lazy from the very beginning. They got a little better but smaller models really only want to have a single task given to them. Mythos at least does work. RIP