SIGNALprophet format
AI features are arriving on personal devices in a state that has no public name yet: shipped, installed, deliberately inert. Major vendors now have a ready line for that posture, and the public has no portable label to push back with. Both gaps are the story.
In June, after WIRED reported that Meta's NameTag face-recognition system sat dormant inside the Meta AI app on tens of millions of Ray-Ban smart-glasses companion installs, Meta's VP of communications Andy Stone posted the line that defines the gap. Independent researcher Buchodi had run the code at WIRED's request and matched a photograph of Michel Foucault. The day after the report, Meta deleted the code. A month later, on the July 8 episode of The Most Interesting Thing in AI, Meta CTO Andrew 'Boz' Bosworth told host Nicholas Thompson how the feature would work: a person met in person, glasses on, name surfaced, or a face the glasses were told to remember with 'OK, this is David, remember this person.' He called it 'a NameTags feature' and said it 'would be a great feature.'
The contradiction is not the story. The story is the category. Code can sit on hardware, in production apps, on tens of millions of devices, with no feature, no disclosure obligation, and a vendor line ready to deploy the moment a reporter arrives. 'Shipped-but-inactive AI' is the reusable label, and Stone's 'doesn't exist' formulation appears to be establishing a playbook vendors may reach for when a reporter arrives. What a vendor owes the public when inert code is already on user hardware is now the question every smart-glasses launch will have to answer.