The most useful screen might be the one that disappears
A message arrives in your glasses. A moment later, you get behind the wheel. The message vanishes from your field of view; the assistant can still answer through audio. In Meta’s patent drawings, the impressive trick is the empty space where the interface used to be.2014
That is a useful way to read the company’s recent glasses filings before Connect. Meta said it would have more to share about its glasses lineup at the September 23 event. The records examined here point to a deeper question than whether another display or camera will fit into a Ray-Ban frame: can a computer on your face become useful without making every moment a computer interaction?19
The applications describe ways to clear the display, limit what a camera passes onward, make conversation easier to hear and accept speech-related input without an ordinary spoken command. They are separate engineering proposals, not a confirmed feature list for the next Ray-Ban product. Their value is in the decisions they make visible—and the questions they give us to ask of a live demonstration.
First, know when to get out of the way
US20260229154A1, published August 6, treats a display as something the glasses can switch out of, while continuing to respond through speakers. Its examples include automatically suppressing the display during driving and asking a cyclist to turn it off. When the relevant condition clears, the interface can return automatically or after a prompt. The broader mechanism also lets the wearer wake the display again.1
This is more interesting than a generic “do not disturb” toggle. The assistant can keep doing work while changing how the wearer receives the result. A visual confirmation can be useful at one moment and a distraction at the next. A display-equipped pair of glasses can temporarily behave like audio glasses.
The original filings let us check how early that idea existed. The January 31, 2025 provisional, 63/752,499, already contains the driving and bicycle sequences. In the corresponding passage of the July 8 provisional, 63/840,575, Meta elaborates the driver-versus-passenger problem: a forward camera could identify a steering wheel, and the determination could draw on camera, audio, location, motion or other sensor data. These are specific additions in the passages compared here; they are not a conclusion that every feature in the later application has a particular priority date.1415
The demonstration worth watching is the transition. Does a useful answer survive the change of mode? Does the display return at the right time? And when the glasses misread the situation, can the wearer correct them immediately? The filing supplies a design for that problem, not evidence that driving with an audio assistant is safe.
A camera that passes along meaning
The privacy question starts before a photograph reaches an app. US20260252168A1, published August 27, puts object detection on the image sensor. The sensor recognizes objects, generates codes describing them and encrypts those codes for an off-sensor processor. In the original Figure 1, a scene containing a car and a dog becomes object codes, then encrypted object codes.212
That changes what the next stage of the system needs to receive. An assistant that needs object context might not need the original pixels for that task. The February 27, 2025 provisional, 63/764,049, already describes onboard detection and an encryption circuit; this is not an architecture inferred from a later marketing description.11
There is still a meaningful tradeoff. Object codes carry information about a scene, and their usefulness depends on the sensor identifying the right objects. This proposal does not show that all camera modes avoid transmitting images, that an entire product is private, or that every assistant task can be reduced to object labels. It gives us a concrete reporting question: for this feature, what exactly leaves the sensor?
And a mute control that software cannot casually undo
A separate family asks what “off” should mean. US20260244794A1, published August 20, describes a hardware-controlled sensor mute state. Its original February 18 provisional, 63/760,047, describes applying the mute state as a single hardware operation, enforcement settings that software cannot subsequently alter except through a hardware reset, and separate secure and non-secure sensor-processing paths.313
The distinction from the object-code design matters. One limits the information passed along while a sensing function operates. The other governs whether sensors may operate and how their processing paths are isolated. The latter disclosure also addresses coordination between privacy indicators and actual sensor operation: a light that says “off” is only useful if it tracks what the hardware is doing.13
Read together, the two families suggest a more demanding way to assess a camera assistant. Ask both what it shares while it is working and what enforces its state when you turn it off. They remain separate proposals; the records do not establish that a current or forthcoming pair of Meta glasses combines them.
Two ways to hear the person across the table
Now move from the road to a noisy table. There are two different places to improve the conversation: at the listener, where competing sounds arrive, or at the talker, before their voice has to cross the room.
The listener-side proposal is US20260255124A1, published August 27 and claiming January 17 and February 24, 2025 provisionals. It uses motion data to infer an auditory zone of interest and selectively enhance sound from that zone. The specification describes a focus anchored to the surrounding world, rather than assuming the desired speaker is always directly in front of the wearer’s face. That matters when you briefly turn your head while continuing to listen to the same person.7
US20260270630A1, published September 10 and claiming provisional 63/767,613 of March 6, takes the talker-side route. The glasses capture their own wearer’s voice and send it to another person’s wearable audio device. This is a direct link from the person speaking to the person listening, distinct from enhancing the listener’s surrounding sound field.8
The comparison produces different tests. For the first, watch whether the system follows the conversation the wearer intends to hear when their head moves. For the second, ask how people establish the link and what happens when the conversation changes participants. A quiet stage demonstration would tell us little about either problem.
The reply you would rather not say aloud
Listening discreetly is only half the social problem. There are plenty of places where speaking an instruction to your glasses is awkward. US20260253591A1, published August 27, explores detecting speech-related activity through sensors in contact with the head or face. Its February 25, 2025 provisional, 63/762,845, discusses electrical muscle signals, contact microphones and motion or vibration sensing.918
The appealing scenario is a private instruction without an ordinary voiced command. The mechanism still depends on physical activity associated with forming speech. It does not read arbitrary thoughts. Mouthing words, whispering and producing detectable activity without audible speech also impose different technical demands; a demonstration should say which one it is doing.
This is the frontier portion of the story. A convincing demo would need to show the permitted vocabulary, calibration, sensitivity to frame fit and the rate of unintended activations. A successful command on stage would not by itself establish an everyday replacement for speech or touch.21
The frame has to carry all of this
Every additional sensing or display function still has to fit into something a person will wear. The most tangible hardware idea in this group appears in US20260237727A1, published August 13: a polymer-electrolyte battery integrated into a temple arm, with a bonded hermetic coating. A dependent claim expressly adds mechanical support of the temple arm.4
The original February 10, 2025 provisional, 63/756,718, makes the proposal easy to see. Its Figure 3 labels a structural temple-arm battery, a speaker, electronics and terraced battery sections. The specification describes tuning the battery’s stiffness, reducing the surrounding packaging burden and shaping cells to occupy curved parts of the frame.1617
The ambition is for the battery to contribute to the structure that holds it, while recovering space otherwise lost to packaging and awkward geometry. The unresolved questions are equally physical: repeated bending, moisture protection, manufacturing tolerances and long-term capacity. The reviewed record does not establish measured gains in battery life, weight, yield or cycle life. It is a strong direction to follow, with much weaker grounds for predicting a product announcement next week.
Two more problems beyond the next demo
The rest of the shortlist broadens the engineering picture. US20260244025A1, published August 20, uses body-coupled communication in eyewear: one electrode contacts skin, while another arrangement can use a transparent lens conductor as a floating ground. Its claimed provisional is 63/760,324, filed February 19, 2025. The June 18 publication US20260172123A1 addresses a related but separate wearable-network problem—adapting communication as body impedance and contact conditions change—and claims five provisionals from December 2024 through July 2025. These are communications proposals, not evidence of thought sensing or of Bluetooth disappearing from a named product.56
US20260177838A1, published June 25 and claiming 63/736,435 of December 19, 2024, tackles prescription display optics. Light from the real world and light making the virtual image take different paths; its holographic optical element incorporates correction for the virtual image. Ordinary prescription support is not the novelty being claimed here. The interest lies in making the display’s optical path work for the wearer as well as the view through the lens.10
These subjects deserve their own treatment. For this preview, they explain why adding a feature to glasses can also create a connection, fit, power or optical problem elsewhere in the system.
At Connect, watch the handoffs
The records give us a more useful checklist than an invented product specification. Watch a visual answer become an audio answer without losing the task. Watch what happens when the glasses infer the wrong context. Ask what the camera passes onward, what actually enforces a mute, and whose voice a listening feature is designed to improve.
If silent input appears, ask what physical action makes it work and how often it fires when it should not. If lighter frames or longer endurance are announced, look for measured results before connecting them to a structural-battery proposal.
The common question is whether the interface leaves you in control of your attention. A computer that is always available has to be unusually good at knowing when to disappear.
Evidence notes, claim details and open questions
What these records establish
Research through September 16, 2026. This pass screened 133 Meta Platforms U.S. application publications returned for June 1–September 16; the latest imported publication week was September 10. A parallel Facebook-name search added no results. Fifty opening priority statements identified a continuation or divisional. Those were not treated as new underlying research simply because their publication date was recent.
The ten publications discussed here claim 16 provisionals, with no earlier U.S. publication found in the retrieved application and continuity records. We treat them as first-U.S.-publication candidates. International publication coverage was insufficient to establish the first worldwide disclosure, so “newly visible” here does not mean proved secret until the U.S. publication date. Provisionals are not independently published application publications; their underlying files can become accessible through related public records. Nor does a priority claim prove that a provisional supports every later feature.
As of 2026-09-16 18:18 UTC, all 479 indexed documents across the 26 selected U.S. files were downloaded. 455 had text; OCR reported 454 completed and 25 skipped, with none pending. Acquisition is not substantive review. This article relies on the specific publication passages, original provisional passages and drawings identified in its source pinpoints. It does not claim a complete review of every appendix, all prosecution events or international file histories.
The family maps below are native patent-platform exports. The map for US20260255124A1 is withheld because the returned graph incorrectly includes an unrelated medical branch. Its publication and two expressly claimed provisionals remain identified above; we have not substituted a reconstructed family tree.
The people panel links the on-sensor application’s signed data sheet and inventor declarations to their original pages. It identifies documented participation; it is not a complete contributor list for all ten families.2223