Uncertainty-Routed Human-LLM | Pangram Labs