Dev gets nearly a small LLM running at 10 tokens a second locally on a $10 microcontroller

28.9M-parameter model runs on a microcontroller costing less than $10 ...