Local Model Routing: Run AI for $0

Local Model Routing: Run AI for $0 (Then Escalate When It Matters) I run two machines. A Mac Mini M4 handles orchestration — talking to users, making decisions, dispatching work. A Mac Studio M4 Max runs Ollama with three local models: deepseek-r1 (5GB), qwq (19GB), and gemma3:27b (17GB). Between...