Apple's VisionOS 27 developer beta gives Siri on device vision, and the bigger story is the broader race to put camera aware AI on faces and wrists, with a privacy cost that has barely started to be priced in.
In Scott Stein's office, a stack of magic trick books sits on a real shelf. When he asked Vision Pro's Siri to look, a glowing orb appeared, then a plain English description: a collection of magic books and props, would you like to know more. That moment, described in Stein's CNET hands-on of the VisionOS 27 developer beta on June 12, 2026, is not really a headset feature. It is a preview of the next decade of personal computing, and the headset is just the first surface it shows up on.
Stein, a CNET editor at large with nearly two decades of wearable and mixed reality coverage, framed the test as a phase change. Persistent camera access turns Siri from a voice assistant into a sensory companion that, on command, describes everything in your field of view. Virtual Paris windows mapped onto real office windows. A wall clock that is not really there. The interface shift is the part that matters: gaze and camera become the new keyboard, and the headset is just the most visible place that shift is landing.
The CNET piece positions VisionOS 27's visual Siri as a preview of camera-enabled wearables more broadly, including smart glasses form factors, and the broader race to put ambient, gaze-aware AI on faces and wrists is the category the developer beta is signaling. That framing is editorial, not a product announcement. The build itself is rough: features are gated, behavior is inconsistent, and Apple has not published a consumer release date or a privacy posture for the new surface in the developer beta.
The privacy cost is the part that gets underweighted in most coverage. An assistant that watches your life on command has to make a constant set of decisions about what is processed on device, what leaves the headset, who is in frame, and what gets logged. The answers to those questions will shape whether consumers actually adopt camera-aware assistants or quietly leave the camera off. Stein's own caveats, that not everything works and that the always-on posture is uncomfortable, are part of the report, not a footnote.
The user experience question of the next decade is not which assistant answers fastest. It is what we choose to let our devices see, and whether the companies building them will let us choose at all.