For ambient AI, the binding constraint is sensor geography: a model that needs to reason about a user's world requires persistent, low-latency access to what that user sees and hears. Glasses solve the physical positioning problem that earbuds and phones address only partially. Big Tech has concluded that wearable eyewear is the distribution layer for artificial intelligence at the consumer edge.
The mechanism behind that positioning is presence versus availability. A phone is reachable; glasses are already in the field of view and audio environment. For an AI inference pipeline that depends on environmental context rather than triggered queries, the difference in sensor geometry drives the architecture.
Critics have a different frame. An always-on camera and microphone array embedded in a glasses frame captures bystanders who provided no consent. That is the source of the "cringe stalkerware" label that critics have attached to the category. The characterization points at a data capture problem that no edge-processing optimization fully resolves: the sensor sees what it sees, regardless of where the inference runs.
Big Tech's position treats wearables as the gateway to AI. The criticism treats the same device as a tool for involuntary data collection. Both descriptions point at the same hardware.