AI agents on smart glasses now work continuously — and people are already delegating instead of controlling
What happened
Researchers built an always-on AI agent that runs on Meta Ray-Ban smart glasses, continuously watching what you're looking at and executing tasks you speak aloud — adding items to shopping carts, transcribing documents, controlling smart home devices. In tests with 12 people in a lab and 5 in the field, the system was faster than non-continuous alternatives, and the behavioral shift was surprising: people stopped treating it like a tool they controlled and started treating it like a delegate they trusted with opportunistic tasks.
Why it matters
Until now, wearable AI agents required deliberate activation — you had to ask them to start paying attention. Continuous perception changes the interaction model fundamentally. People stop thinking about 'asking the AI' and start thinking about 'the AI noticing and handling it,' which means the cognitive friction of delegation drops below the friction of doing it yourself. The deployment study is the actual signal: five people, real conditions, and the behavior shifted from directed control to ambient delegation. This is how technologies move from gadget to reflex.
The signal
Whether people wearing these systems over weeks or months drift toward relying on opportunistic task execution (the glasses noticed I had a receipt, so it extracted the data) versus directed control (I told it to extract the data), and whether that shift causes any attention or safety problems in real environments outside the lab.