Running Ollama on a 15W CPU sounded ridiculous until I got it working with decent results
…That's fairly underwhelming for a local LLM setup, especially the memory, since we're going to be CPU-bound and that RAM is super-slow compared to DDR5 and a discrete…