Open with the filing pattern, then name who owns it. Aggregate recent spatial-computing grants by CPC subclass and G06F 3/013 — eye-tracking input — is where the records pile up. This is not an accident of classification; it is the companies telling you, through where they spend prosecution dollars, that gaze is the contested interaction primitive of head-mounted computing.
Apple's lane is gaze-as-everything: control input (US12645292B2, head-mountable display, June 2, 2026), rendering budget (US12650731B2, distributed foveated rendering, June 9, 2026), and focus (US12647549B2, gaze-driven depth of field, June 2, 2026). Three grants, one signal axis.
Magic Leap's lane is fusion. Its grant US12651420B2, "Transmodal input fusion for a wearable system" (issued June 9, 2026; assignee Magic Leap, Inc.), classifies to G06F 3/012 and 3/013 alongside gesture and voice — combining gaze with other modalities into one interpreted intent. Where Apple treats gaze as the spine, Magic Leap treats it as one input among several to be fused. That is a genuinely different architectural bet, visible in the CPC mix.
Qualcomm sits on the capture side. US12647692B2, "Foveated imaging" (issued June 2, 2026; assignee QUALCOMM Incorporated), applies foveation to the camera pipeline rather than the display — the silicon vendor staking the sensor-and-ISP end of the same idea. As a component supplier, Qualcomm patents the part it sells; the lane fits the business model.
The white space worth noting: there is far less granted density on what to do with gaze data beyond UI and rendering — gaze as a health, attention, or authentication signal is comparatively thin in these results. White space is a strategy tell. The companies are, for now, fighting over gaze-as-interaction. The reader watching for the next front should watch whether the cluster starts spreading into A61B (medical sensing) — that would signal gaze moving from how you control the device to what the device learns about you.