A better headset has to do more with less on your face
A headset can show a sharper world and still feel like a computer strapped to your head. Meta’s filings keep returning to the things a resolution specification leaves out: where the work happens, where the heat goes, how the cameras represent nearby objects and whether the optics put the image where your eyes need it.1653
The useful “what’s next” story is several connected engineering problems. The records do not establish that their solutions will appear together in a named next-generation Quest. Quest 3 supplies a familiar pancake-optics baseline; newer publications and research prototypes show possible directions beyond it.1314
Move the work, not just the battery
US20260169761A1 places graphics processing in a handheld intermediary device and presents the result through a separate head-wearable device. The companion is part of the computing arrangement. Its disclosed role goes beyond supplying power: it processes shared information and helps produce the artificial-reality experience.17
The August 14, 2023 provisional already describes an intermediary generating an AR representation and instructing the head-worn display. Its first dependent claim even permits a companion without a screen. The June 18, 2026 publication is therefore a continuation of an older direction, not evidence that Meta first conceived of companion processing this summer.178
The file history now gives that companion a more specific job. The September 1, 2026 allowance includes a three-device arrangement: a handheld processor, a distinct wrist device and a headset sharing image data. When the headset’s thermal level calls for moving back-end tasks, the handheld takes them on. The headset presents the image and receives control through commands from the wrist device. The allowed claim combines those roles; it is not a claim to any headset with an external processor.15
That distinction mattered in Meta’s July response. The applicant argued that offloading because of the headset’s thermal state was different from merely observing that offloading could reduce heat. It also insisted that the wrist input device and handheld processor were separate devices. This is a concrete direction to watch: the display, the work and the input can live in three different places.1615
Which camera jobs belong on the companion?
The more specific processing split appears in US12368967B2. Camera data starts with Bayer processing on the head-mounted device. A controller decides whether later YUV processing should remain there or run on a companion, considering battery power or application state and checking the communications channel before transfer.2
This is a useful way to judge an offloading proposal: identify the exact job, what data crosses the link and the conditions under which moving it makes sense. A remote processor may reduce work on the face, but the transfer still has costs. The patent supplies a partitioning strategy; it does not supply a verified retail-headset battery-life gain. The application was filed March 8, 2023 and granted July 22, 2025, with a continuation in the retrieved family.2
The continuation, 19/249,847, shows why “move processing off the headset” is still too general. Its August 26, 2026 final action rejects claims 1–6 while allowing other combinations. One sends both raw camera data and locally generated statistics to the companion when battery power falls below a threshold. Another ties transfer of partially processed images to an application capturing video—a condition added to independent claim 15 in June. These are two different choices about what crosses the link and when. The examiner accepted those combinations while continuing to reject the battery-triggered processed-image route in claims 1–6. The application has allowed subject matter, not an issued continuation patent.1817
An accessory can become another way out for heat
US20260239588A1 describes selecting a heat path when the device detects thermal contact with an accessory. Its published claim combines conductors, contact detection and switching between paths. The accessory is part of heat management, rather than necessarily another source of electrical power.6
The July 11, 2022 provisional frames overheating as a limit on sustained operation: processing may have to slow or stop. The current continuation was filed December 29, 2025 and published August 13, 2026. My inference is that this family belongs beside companion computing because both address the thermal burden of wearing the device. The record does not prove that they are one integrated product design.612
Make the nearby object look right where you are looking
A passthrough camera is not located at the eye. Reprojecting its image to the eye’s viewpoint requires an estimate of scene depth, and errors can make nearby objects look misplaced. US12423916B1 uses gaze and vergence to identify the object the wearer is attending to and derives a separate distance from each eye to that object.5
The system then builds a constant-depth mesh for each eye and warps the corresponding camera image. The engineering tradeoff is selective: improve the geometry for the attended depth, rather than claim perfect reconstruction of every surface in the room. This application was filed March 21, 2024 and granted September 23, 2025. The refreshed source record returned no related applications, which the family section states explicitly.5
Move a reflector inside the folded optical path
Pancake optics make an optical path fit into a thinner package by folding light between elements. Meta’s US12517343B2 asks what happens if one of the reflectors can move. Changing its axial position changes the optical power and focal setting of the arrangement.3910
The original October 7, 2022 provisional shows alternatives: translate the partial reflector or translate the reflective polarizer. The drawing makes the proposal easier to understand than the title does. This is physical optical adjustment; a change in rendered sharpness alone would not be the same mechanism. The nonprovisional was filed December 12, 2022 and the patent issued January 6, 2026.39
A holographic image still has to reach a moving pupil
US12607855B2 goes further into display architecture. A spatial light modulator changes a light wave so that the resulting optical field forms the image. That is a different proposition from using a passive holographic element merely to redirect light. The hard practical question is where the wearer can put an eye and still see the image.4
The family addresses pupil positioning and replication. The issued claim includes beam redirection toward successive locations in the eyebox and a spatial filter that controls unwanted diffraction orders. The November 17, 2022 provisional already discusses replicated light, spatial modulation and following the pupil. That establishes a research lineage; the precise issued combination should still be distinguished from the earlier disclosure.411
This is the more ambitious frontier in the set. It should not be presented as a confirmed optical engine for the next consumer headset. Publication and grant tell us what entered the public patent record, not whether a display has met the cost, brightness, efficiency and manufacturing requirements of a product.4
Do not build an imaginary headset from two prototypes
Meta’s own 2025 research presentation makes the tradeoffs unusually visible. Tiramisu emphasizes high angular resolution and brightness but has a narrow field of view. Boba 3 explores a much wider field of view with different display and optical compromises. The company presents them as research prototypes.14
Taking Tiramisu’s sharpness, Boba 3’s width, a companion processor from one patent and a moving lens from another would create a compelling specification sheet. It would also describe a machine these sources do not establish. The more useful synthesis is a set of research directions: move computation, manage heat, improve attended-depth passthrough and explore better control of light reaching the eye.1416534
Evidence notes, claim details and open questions
Research coverage and what comes next
This selection covers six focus publications and 15 retrieved U.S. application records, including four provisionals. Eleven native family diagrams were retrieved across the combined Meta and Apple research; Meta accounts for five diagrams and one application with no related records returned. The search is targeted, not a complete worldwide portfolio census.
All indexed document codes were requested for each member. The original focus publications and accessible provisional specifications, claim sheets and drawings are archived locally. A local OCR pass made selected provisional pages readable while platform OCR continued. The long intermediary-device provisional was read selectively: its opening specification pages and claim sheets, not all 105 specification pages.
The article is a technical research draft. This pass reconciles the companion-device allowance with its amended claim 1 and reviews the ISP continuation’s June claims and mixed August action. Other selected responses were read for context; every amendment across all families has not been reconciled. Product evidence connecting these mechanisms to an announced device remains a separate question.
Acquisition checkpoint, September 14, 2026 at 9:57 PM Arizona time: 642 of 642 indexed documents downloaded across 15 applications; 604 OCR-complete and 38 skipped, with none pending, processing or failed. The skips comprise 23 specification/abstract records routed to publication text, six actions routed to another text archive, and nine records without a stated skip reason. The August ISP action’s archive fallback did not return text, so its original PDF was checked directly and locally OCRed. Queue completion does not mean every page has been read.
International coverage gap: the refreshed platform records identify PCT/US2024/042241 and PCT/US2023/079961, but their filing dates are missing upstream, so the native SVG omits those nodes. The manifests flag this limitation. Those PCT histories are not included in the U.S. download totals and remain to be acquired and reviewed.