My local AI ran on the CPU for a week and I didn't notice

My first Ollama setup felt fast. One evening, one model, one question, a clean answer streaming back...