DeepSeek's Harness team released "Super-agent Design Exploration: Toward a More Autonomous Agent System," introducing an engineering-practice-oriented architecture designed to overcome limitations in current agent frameworks. The company identified key pain points in existing single-session Agents, including the need for manual task switching, limited concurrent session tracking (typically 3-5 active sessions), single-session blocking, context confusion, and task interruption when local machines go offline. The proposed Super-agent architecture is characterized as continuous, efficient, proactive, and persistent, built on five core components: a two-layer Router and Worker structure, event-driven design, human-oriented interaction, separation of compute from state and environment, and a Watchdog mechanism. In this framework, the Router manages human interaction, maintains global context, and dispatches tasks, while Workers execute concurrently in isolated sessions—reporting status and results back to the Router rather than responding directly to users. Additional mechanisms include Trigger event filtering, Inline & Spill, chained recall, speculative execution, asynchronous ReAct Loop, and Watchdog monitoring.