When an IoT platform composes automation rules across Things, rules can conflict. Two rules targeting the same device with opposing states. One rule's output triggering another in a cascade. Alert storms that flood operators with notifications. I wrote ExMaude—an Elixir library that supervises the Maude rewriting-logic interpreter—to catch these conflicts before activation. In this simulation, ExMaude checks the same rule representation the runtime evaluates and returns one of three results: conflicts, clean within the model, or unverified. Four modelled conflict categories. The live demo compares observe and enforce modes under the same synthetic load. This talk covers the problem (with a published smart-home conflict pattern), the solution (Maude rewriting logic explained for Elixir developers), and the integration (a narrow deployment gate in a Phoenix/BEAM simulation). BeamLens and Codex or local Ollama explain a bounded telemetry snapshot; Maude remains the deterministic decision maker.