There is an awkward compute problem hiding inside AI glasses: the agent should always be available, but cloud reasoning should not always be running.

NewEyes AI from @CollovLabs takes that problem directly.

It is now on Meta glasses, with lightweight processing handled locally and deeper reasoning moved to the cloud when a request actually needs it.

That separation is the part I find more interesting than the glasses themselves.

A visual agent can potentially observe far more information than a user would ever type into a chatbot. And that becomes useful in very ordinary situations.

You could be cooking and ask about what is in front of you, putting together an outfit and getting help based on what the glasses can see, or working through some small task without stopping to pull out your phone.

The interface is basically: look, ask, act.

For this kind of product, the hard part is making AI available throughout the day without treating every visual moment like it deserves a full cloud reasoning pass.