From Explainability to System Language: Why AI Accountability Requires a Formal Execution Vocabulary | Synapse