Meta Muse reveals six subagent worker types

Consumer AI agents are rapidly shifting away from monolithic chatbots toward orchestrators that manage ephemeral, specialized subagents within sandboxed environments. Muse favors on-demand instantiation of task-focused subagents over persistent personas, mitigating context drift, reducing token overhead, and ensuring tool boundaries remain clean. Dedicating an explicit Browser Operator agent to navigate authenticated web flows, form completion, and purchases aligns with Meta's Sentinel VM architecture designed to isolate execution risks. Including a built-in Workflow Orchestrator indicates that Meta is bringing developer-grade DAG pipelining directly into mainstream consumer mobile experiences, while demand for hooks, tools, and subagents underscores that extensibility and custom skill ecosystems are the next competitive frontier.