Original · Unique Research · 2026-04-24
Editor's note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the opening essay, all section headings, the full guest roster, and the complete panel transcript. Hardware specifications, sales figures, market-sizing, battery life, weight, and pricing claims are source or speaker statements, not independently audited findings. Company, personal and work titles are transliterated where official English forms remain unverified. DigiWindow, VOCCI, Even Reality, and other product names are retained as the source's naming. The "trillion-market" and "last technological revolution" rhetoric are speaker opinions, not established forecasts. The source is dated April 24, 2026.
Global Unique Awards
These People Making AI Hardware in the Greater Bay Area Discovered the Phone Is Not the Competition at All
AI Agent is clearly regarded as the soul of the next era—so why does it become maladapted the moment it is put into a hardware "shell"?
"
In 2026, the prerequisite for making AI hardware well is precisely to "step back."
Globally, more than a billion people wear prescription glasses every day. But after hardware fanatics stuffed AI large models into the frames, an extremely awkward truth emerged:
Users simply cannot bear to wear them.
To add a camera and a screen that can read WeChat, AI glasses easily exceed 40 grams in weight, and the battery runs out after an hour of being on. The result is that this "most powerful brain," trained at a cost of billions, has in reality been reduced to an "advanced toy" occasionally used to take a photo.
AI Agent is clearly regarded as the soul of the next era—so why does it become maladapted the moment it is put into a hardware "shell"?
At a recent "Hong Kong · Global Unique Awards" trends roundtable, four top entrepreneurs deeply involved in AI hardware, XR, wellness, and voice interaction held a fierce debate titled "The Soul in the Shell." They uncovered a counterintuitive business logic: in 2026, the prerequisite for making AI hardware well is precisely to "step back."
To Make AI Disappear, They Built a 1-Gram Screen
AI hardware has always been unable to escape an "impossible triangle": more features, longer battery life, lighter weight.
If you want the screen on, the battery has to be bigger, and the glasses become heavy enough to break the bridge of your nose; if you want it as light as a feather, then all you have left is a microphone.
"The ultimate AI hardware should make people feel nothing." Jiang Dawei, General Manager of Greater China at GYGES LABS, chose an extremely extreme solution to this problem.
This is a team composed of Stanford PhDs. Instead of obsessing over the extremely power-hungry traditional optical waveguide lens, they "hand-built" the world's smallest display system, DigiWindow—weighing less than 1 gram. Stuffed into glasses of less than 30 grams, with power consumption under 30 milliwatts, the battery lasts 3 to 4 days.
This is a very instructive retreat: it no longer tries to give you a dazzling virtual world, but only prompts you with the 100 most important pieces of information when you need them. For example, during a business negotiation, AI actively hears "Hong Kong population," then silently projects the answer at the edge of the glasses: Hong Kong has a population of 7.5 million.
When you are not using it, it is like a normal pair of framed glasses with a 400-degree prescription lens. This is precisely the cleverness of the hardware: it does not change user habits, it just secretly adds a little magic.
"Don't Try to Replace the Phone—Be the Phone's Little Brother"
Many people question: since the phone has such strong computing power, can record audio and translate, why should we spend an extra few thousand yuan on a standalone AI voice recorder?
Hua Kun, founder of Wavenote, gave an unusually clear-headed positioning: "Because the phone is the big brother, and the big brother is too busy."
The phone has to take calls, scroll videos, reply to WeChat—it cannot handle a two-hour deep recording session or high-frequency foreign-language translation. One phone call comes in, and the recording may be cut off.
This is why traditional voice recorders still sell tens of millions of units every year, and after adding AI large models, this market has welcomed a ten-million-level explosion. In the eyes of hundreds of millions of heavy office workers globally, what they need is not a device that can do everything, but aparanoid tool that "records from far away, doesn't make you wait for transcription, and is absolutely reliable."
The same emotion appears in the wellness track. Li Shuting, CCO of Dongfang Qiming, brought an "AI ring." It not only monitors heart rate, but behind it runs a 32-billion-parameter Eastern culture large model.
Imagine: you have just landed after a transnational flight, and a ring gently reminds you: "Today you are exhausted from travel; it is advisable to sit quietly for three minutes."
This kind of emotional comfort and noise-canceling moment can never be replaced by the phone's cold push notifications. Hardware is becoming a soul container that provides emotional value.
The Overlooked Disruptor: When Hardware Learns to "Actively Eavesdrop"
If glasses and voice recorders are efficiency tools, then what exactly has AI brought to the XR (extended reality) industry, which has always been lukewarm?
"In the past, AR and VR could only provide visual 'display.' But what AI brings is 'interaction' with the real world." An Shanzi, founder of Zhuijiao Tech, who has been in the XR industry for 11 years, pointed out a core turning point.
Hardware in the past was passive—you press a button, and it gives you a response.
AI glasses now are active. An Shanzi mentioned a pair of Even Reality glasses that sold explosively overseas. To pull the weight down to an extreme 38 grams, it directly cut the external speaker and camera, keeping only the microphone.
Why does it sell well? Because it enters an "Always On" state. It listens to what you are doing for long periods, actively captures keywords during conversations, and feeds you information in display form. Hardware has changed from "a machine waiting for instructions" to "a first-person Agent that understands the world."
This pragmatic design is killing it in overseas markets. Compared to MR headsets costing tens of thousands of yuan, weighing 200 to 300 grams, and facing a relatively cold market this year (such as Vision Pro), overseas consumers are extremely willing to pay for lightweight AI glasses at US$600–800.
Maximize wearing comfort and penetrate one specific scenario thoroughly—that is the wealth code for going overseas.
If Starting Over, Which Card Would They All In?
What will future agent hardware look like?
It is the minimal voice entry that Hua Kun envisions—one that can understand "whispers," free the hands, and solve privacy problems. It is the AI ring in Li Shuting's hand that can accompany you 24/7 and reduce costs and increase efficiency for nursing homes. It is the 1-gram display from Jiang Dawei's team that can be clipped onto a hat or even a surgeon's scalpel.
Or perhaps it is the XR headset that An Shanzi looks forward to—one combined with Fei-Fei Li's "world model," where you only need to speak in natural language to generate a real space like a creator.
No matter how the forms vary, the Greater Bay Area hardware entrepreneurs have reached a consensus: in 2026, hardware is no longer a carrier of information, but your "digital avatar" in this physical world.
In an era that is not short of shells, whoever can first put in the "soul" that best understands the user will get the ticket to the next trillion-dollar market.
More Dialogue Details
Hong Kong · Global Unique Awards Trends Roundtable Panel
Theme: The Soul in the Shell: Agents Packed into Devices and the New Business Opportunities of AI Hardware
Guests: Dongfang Qiming — CCO (Chief Customer Officer) — Li Shuting, Wavenote — Founder & CEO — Hua Kun, GYGES LABS — Greater China Business General Manager — Jiang Dawei, Zhuijiao Tech — Founder — An Shanzi
Moderator: World Manager — Tech Vision Host — Kang Zhengzhong
Kang Zhengzhong: I am from Global Sources. We have a business management media called "World Manager," and I am also in charge of the "Tech Vision" column. Our panel is called "The Soul in the Shell." Actually, we just discussed a lot about AI software, including some applications, but in our Greater Bay Area, you cannot separate from hardware, right? So this topic should actually be approached from two angles. The first is to discuss "shell" and "soul"—the shell is our hardware, and the soul is of course the now-commonly-mentioned Agent. In the past few months, actually the two most discussed things are Physical.ai and the so-called OpenClaw. Everyone says, wow, AI Agent is so powerful, robots are coming—but what conclusion do you get when the two are combined? Actually, many times we are also exploring this. So before entering this topic, we have invited four top entrepreneurs in this industry. Let us first ask them to use about one minute to introduce themselves and the business behind their companies. Mr. Jiang, would you start?
Jiang Dawei: My name is Jiang Dawei, from GYGES LABS. I can tell you a story about how our company got its name GYGES—it comes from a Greek legend. Plato wrote a book called "The Republic," in which a shepherd found a ring in a cave. After taking the ring, he gained a magical power to make himself invisible, and eventually he became the king of this country. Then that shepherd was named Gyges, and we are called GYGES Labs. And today I also brought our company's product—you can all see this ring I am wearing. This ring can control my glasses. I can also give a brief demonstration. Here, let me put in an advertisement for Global Sources: I can see a screen—Global Sources is China's oldest international B2B cross-border e-commerce platform, founded in 1971, formerly known as Asian Sources, headquartered in Hong Kong, listed on Nasdaq in 2000.
Hua Kun: Hello everyone, we are Wavenote. We actually have three products in different forms, because the voice recording track is still very large. We started in 2024 to discover the combination of large models and this, and the potential here, because indeed a lot of offline voice information was not well utilized in the past. Of course, it seems like many people are doing this, but actually I think not many do it well. I think this is mainly a "software-defined hardware" track—more about software bandwidth. Actually, what we are building is an Agent. Of course, the hardware itself must be done well, because the first thing users buy is actually a piece of hardware.
Kang Zhengzhong: Mr. Hua, is your hardware form currently earphones? Or?
Hua Kun: Of course we will adapt to user needs. For example, business people still prefer this portable card form, but for schools and universities, they prefer this wearable or this magnetic-attachment method.
Li Shuting: I am Li Shuting, CCO of Dongfang Qiming. Dongfang Qiming is an "AI + Eastern culture" technology enterprise. We have currently collected nearly a thousand classical texts, such as some classic ancient texts like the "Huangdi Neijing" (Yellow Emperor's Inner Canon), which we have digitized. Actually, in plain terms, we take the wisdom of our ancestors from thousands of years, translate it with AI, and put it into our AI products. Our main products now include the Dongfang Hui Ring and Dongfang Hui Eye, which are also our AI rings. Our current core technology is a nationally-filed 32-billion-parameter Eastern culture large model. Our wish is also to let Eastern wisdom be packed into devices, so that everyone can be more exposed to Eastern culture in daily life.
An Shanzi: Hello everyone, I am An Shanzi, founder of Zhuijiao Tech. I just came out of my previous company and am now in a stage of starting my own business. I have been following this industry for about 11-plus years, so it is equivalent to walking with the entire XR industry to where it is now. Now we see that the form of glasses—like what GYGES LABS is wearing—is already in a very thin and light state, so we can see that technology has made great progress. How to put it? We are an industry observer and also a resource connector, so if everyone has needs, you can also contact us.
Kang Zhengzhong: Thank you, Teacher An Shanzi. Let us return to the theme of today's panel. Actually, it should be interpreted from two perspectives: "shell and soul." The shell of course refers to hardware, and the soul I just mentioned should be our AI Agent. The first perspective, let us talk from the angle of technology trends. First, I would like to ask Mr. Jiang: when it comes to this form of AI, AI glasses are definitely an unavoidable must-option. Many people think AI glasses are the ultimate form of AI hardware. Of course there are different views—for example, Mr. Hua and Ms. Li may have different angles. But generally for AI glasses, we usually say it is the first-person perspective of an Agent. So what do you think is the ultimate form of AI glasses—pursuing the ultimate feeling of disappearance, like Teacher An Shanzi, making the wearer feel nothing, or pursuing a stronger sense of interaction?
Jiang Dawei: Because the word "ultimate" is used here, right? The ultimate is definitely still making people wear it comfortably. Honestly, I also see many friends here wearing glasses. In China, probably about 40% of the population wears glasses. Globally there are 8 billion people—not to say half, but close to half may wear glasses. Because I make glasses, we invented this glasses last year. Actually, looking at that screen, this display is our company's patent, because the founding team includes PhDs from Stanford. They hand-built the world's smallest display system in the lab. We have been shouting that slogan—world's smallest display system DigiWindow—it weighs less than one gram. Put this less-than-one-gram display into my glasses, and everyone can imagine, its weight is definitely very light. Because it is not displayed on my lens, so when we communicate together it is not awkward. If it were on my lens, everyone could see it too, right? Let me take it off and show you. At this distance, everyone almost cannot see it, but I can see it. My phone is in my bag. For example, if there is something important at this moment, I can see it—who is looking for me.
Of course, picking up on what Teacher Michael just said, the ultimate is definitely still making everyone feel nothing and wear comfortably. But right now everyone seems to have high expectations for AI glasses interaction, such as Apple's Vision Pro. Our CTO, one of the Stanford-graduated PhDs, when he went to Apple, was on the Vision Pro team. Regarding the current sense of interaction, I personally think everyone may have other ideas. Vision Pro may have a stronger sense of interaction and stronger manipulability. Besides that, the best-selling ones now, like Meta—you cannot say its interaction is very strong; it just added a function to the original product, which is the camera function. I personally understand, for example, in places like California where the light is strong, there are already many Ray-Ban users. They wear sunglasses when going out. If I add a small camera function for you, would you be willing to buy it? Then I would definitely be willing to buy it, right? Or my original sunglasses were already broken, I add a camera, and when I go out and see a scene I take a photo—I feel good about this.
But adding more functions, such as object recognition—of course object recognition is now gradually being done better. And other things like payment, ordering me a coffee, I want to ride a shared bike—these slogans can be shouted, but the current sense of interaction may not be strong. When theuser experience is not strong, many consumers buy it and then may return it. Here let me briefly boast about our product: the glasses we jointly released with Halliday were the best-selling AI glasses in Kickstarter crowdfunding history last year. On the first day of sales we sold nearly 2,000 units—in one day. This is not me bragging; everyone can look it up. On the first day we sold US$1 million, and over the more than 40 days of crowdfunding we sold close to US$4 million.
Because we only added a very, very small change to the glasses—that you originally wear glasses anyway, Michael also wears glasses, so if I add a display on your glasses, would that work? The weight may only add one gram. Of course, this display also needs PCBA and a battery, right? But the entire glasses weight is still maintained at under 30 grams. My glasses are 28.5 grams. That means I can wear them and last a whole day. And because its power consumption is also relatively low, just like the earphones we usually use, power consumption does not exceed 30 milliwatts, so I can last three or four days on a charge, unless I am using it continuously. So the question just mentioned—the future ultimate is definitely nothingness. Everyone who wears glasses hopes to be more comfortable; if it is too heavy and presses on the bridge of the nose, it is uncomfortable. The ultimate feeling is definitely nothingness, but can we wait until that day? I believe we can. The weight of the earliest first computer should have reached 30 tons, and today's laptops are so light. Whether future glasses can reach that day, I don't know how long it will take, but it will definitely be achieved.
Kang Zhengzhong: So Mr. Jiang probably means that first, the mechanical principle of glasses is to make them comfortable to wear, then find somesegmented scenarios to add functions—whether it is the camera or display technology just mentioned. Of course, I have a follow-up question: AI glasses have always had the "impossible triangle" reported in Teacher An Shanzi's articles, right? You want the weight to feel comfortable, you want to add functions, including battery life and these things—how do you make trade-offs?
Jiang Dawei: Because our glasses have not been developed for long, this impossible triangle—of course Teacher An Shanzi is very familiar with it and has paid a lot of attention. We are considered to have done slightly better in this impossible triangle. Why is it called the impossible triangle? It is that everyone wants glasses to have many functions and long battery life. If you have many functions, for example I add a camera, I also need to display the picture. The original optical waveguide display picture is on the lens. The principle is through an optical engine next to it, just like the projector used in meetings, the brightness needs to reach hundreds of thousands of nits, and after multiple refractions it reaches the lens so you can see the picture. Indoors it may be relatively clear, but if the outdoor light intensity is high, your brightness needs to be very high. Then doesn't your battery need to be bigger?
Kang Zhengzhong: Yes.
Jiang Dawei: You cannot have it run out of battery an hour after turning it on. Once the battery gets bigger, the glasses become heavier, and you don't want to wear them. Wanting many functions, long battery life, and light glasses—that is the impossible triangle. Our weight is under 30 grams. Ordinary frames between 20 and 40 grams are relatively comfortable. And our battery life is about three to four days, because the power consumption is very low. There is also our wearing method which is relatively traditional. My lens is 400-degree prescription, the lens can be taken off, does not affect the back. And the lens above can adjust prescription—it supports people with myopia within 1000 degrees. It is by pulling that distance, a bit like an SLR camera and telescope, adjusting the diopter. So it is considered to have completed the foundation of breaking through this impossible triangle. Of course, we also hope to make it smaller and lighter. If AI hardware and software develop one step further, we may find a better balance point.
Kang Zhengzhong: The second question I would like to ask Mr. Hua: there are actually many large manufacturers' software that does recording and translation. In the situation of 2024,targeted at different scenarios there are different hardware entry points. Where is your main technical barrier? Can you share?
Hua Kun: Actually, many friends ask this question: why can't I use my phone or software? Indeed you can, so this group of users is actually not the main users of hardware. But recording equipment has existed for many years. I think many AI hardware products are doing "the phone's little brother," because among hardware the phone is the big brother—everyone carries one. But because this big brother is too busy, he has many things he is not good at or unwilling to do. For example, if you want to record a two-hour meeting, one phone call comes in and it is cut off. So actually many heavy users—globally, this is also a population of hundreds of millions. Traditional voice recorders alone have a volume of tens of millions. Under the value of large models, many people's original potential needs have also been stimulated. So we can see that AI recording products currently have global sales still at the million level, with a lot of room for penetration.
Speaking of barriers: first, hardware itself is a layer of barrier. Because there is such a group of heavy users who use it frequently, for long periods, and have relatively high quality requirements—they will choose to buy professional AI hardware. This track, compared to the AI glasses Mr. Jiang makes, is a larger track—Alibaba and Meta are both doing it; but the recording track is relatively small, suitable for startups. Second, the hardware experience still has many problems. For example, battery life—our product has been optimized quite well and can last a week, users have no battery anxiety; also, how to record farther, can the last row be recorded clearly; how to make the transmission experience better; and the combination with large models. Right now many are still processed in the cloud, and users still need to wait for ASR transcription. The ideal situation is that after recording, users do not need transmission or waiting for transcription, and can immediately view it. We are also researching on-device models. Right now on-device can do limited things, but for some short recordings, we can achieve local transcription combined with the cloud, making the user experience the best. There is still a lot that can be done here.
Kang Zhengzhong: My understanding is that first, we have never thought of so-called translation hardware replacing the iPhone, replacing the phone, right?
Hua Kun: No no, we are a supplement to it, satisfying some heavy users' use in professional scenarios.
Kang Zhengzhong: Then, as you said, finding somesegmented scenarios, such as heavy users like business people, there is actually enough market to support this category.
Hua Kun: They actually have many needs that are not well solved, because the phone is not good at or has not focused on doing this.
Kang Zhengzhong: Then the third question I would like to ask Ms. Li. My understanding is that you belong to the wellness track, right? Because in the wellness track, due to the specific population, it actually pays more attention to so-called functionality. When making wellness-track hardware, do you think giving hardware more is to let it understand the specific population's emotional value, or toachieve the feeling of life's warmth in the hardware form—which is more important?
Li Shuting: Actually, I think understanding the user's emotions is the soul, and its hardware is just its shell. If you have this soul, then it will naturally have a sense of warmth—not just a cold circuit board. Some traditional wellness hardware just gives you real body data feedback, for example your current heart rate is 90. In our product, we will combine the "Huangdi Neijing" with your current body state to give you some prompts. Let me give a real customer feedback example: he often flies transnationally. He said his most touching moment was wearing our ring, and at the moment the plane landed, the ring would give him a very warm reminder: "Today you are exhausted from travel; it is advisable to sit for three minutes." He said those three minutes were his most noise-canceling moment. The viewpoint we want to output is: Eastern wisdom is not teaching everyone how to go faster. What we want to teach everyone is when to slow down, letting the body be in a relatively good comfortable state.
Kang Zhengzhong: That is, the large-model capability for specific scenarios is your technical key, or core competitiveness.
Li Shuting: We have things like solar-term wellness reminders, and daily ones—for example, today at 3 PM is the small intestine meridian's active time, suitable for drinking more water or resting and adjusting. You can treat it as a wellness old friend who understands you best, by your side, giving you comfort and emotional value.
Kang Zhengzhong: That is, you may have some core small-model training data for specific scenarios, and the hardware is just an entry point—maybe a ring, or other carriers.
Li Shuting: Any carrier works.
Kang Zhengzhong: Then the last question about technology trends, I would like to ask Teacher An Shanzi. In the past, in the XR industry, everyone felt XR was a display technology. Today, with Agent as the soul, how does this display technology become an intelligent space that interacts with the physical world? How does it help hardware enterprises transform?
An Shanzi: Mainly like this. Whether AI glasses, AR glasses, or MR headsets, the core function is to bring you a more immersive display effect. AI, starting from 2023 and 2024, has had a very obvious push on our field—whether from the capital level, product level, or market environment. Last year, AI glasses sales reached about 8 million-plus pairs, more than tripling compared to the year before. We forecast this year will reach 10 million-plus units. Many times everyone buys because of AI, but you will find that the value AI generates in current products is not particularly large.
Of course there is a premise: the user base of glasses wearers is very large. Among 7 billion people globally, 2.6 billion wear glasses; among 1.4 billion people in China, 700 million wear glasses. AI glasses are priced roughly from 2,000 to 3,000-plus yuan; overseas high-end co-branded models, such as Google with GUCCI, Meta with Prada or Ray-Ban, may cost several thousand. Overall purchasing power still matches.
About what has AI brought to glasses/headsets? AR/VR gives you display; AI brings interaction with the real world. Divided into several levels: first is "perception," perceiving the real world through cameras and sensors. Second is "understanding"—for example, Even Reality's glasses, you can directly ask it what scene I am in, and based on the camera and positioning it recognizes and tells you you are at the Global Sources exhibition. Third is "decision-making," providing information to you. I recently tried Even Reality's glasses. To achieve extreme lightness, it removed the speaker and camera, only keeping the microphone. It can listen to what you are doing for a long time, and during communication it actively captures keyword information and provides it to you in display form. This lets AI change from purely passive "human issues instructions" to an active form. Finally is "execution" and "evolution"—AI will understand you in the process of continuous learning.
Why do we always think AI glasses are a very good carrier? Because based on the first-person perspective, it understands the world the same as you—hears what you hear, sees what you see. Of course, privacy issues are a concern for everyone, but to obtain convenient and intelligent AI, there may eventually be a balance. AI glasses are good because we have always wanted to achieve an Always On state. Right now it is still relatively difficult because of the impossible triangle. We have now solved the wearing comfort problem from bulky to thin and light. Next is the progress of the underlying supply chain, such as displayeffect, battery energy density, and optical progress. So now glasses forms are blooming everywhere. There are ones specifically for the blind, for cycling sports, for translation, and also ones like Meta Ray-Ban wanting to do spatial computing combined with eye tracking and myoelectric wristbands—this form may be the product that truly replaces the phone in the future.
Kang Zhengzhong: Let us continue to the next perspective of business and globalization. Ask Mr. Jiang: when discussing the smartphone iPhone moment, besides the interaction capability of the Apple phone itself, there was also something very important—its ecosystem. Now making AI glasses, companies like Apple and Meta have their own ecosystems. In the process of commercialization and globalization, because we know you are a Singapore brand with the advantage of China-Asia supply chain. Is your core selling point selling a more convenient pair of glasses, or selling an Agent for a specific scenario on-device?
Jiang Dawei: Our company is headquartered in Singapore, with a team in Silicon Valley developing proactive AI. As Teacher An Shanzi just said, during communication it activates AI, which proactively listens to your chat. For example, hearing "what is the population of Hong Kong," it will give you an information prompt. Besides making hardware, you must base it on consumer usage scenarios. Our current glasses audience is more lawyers, business people, salespeople. For example, when I meet a client, frequently taking out my phone to check messages is very impolite, and opening a laptop to chat facing him is also not feasible. Is having a glasses a bit more feasible? We hope toequipped with thisminimal display DigiWindow into more scenarios—maybe glasses, hats, helmets, a watchmaker's magnifying glass, or a surgeon looking at data during surgery. Any product isequipped with in a scenario.
We also will not be surrounded by Apple and Meta, because the products we make are different. We want to make glasses that can be worn all day, with convenient prescription lenses, not bulky. A small capability is letting you see phone information, so as not to miss the important 100 messages, relieving the anxiety caused by phone vibration. AI glasses are popular because they let people better free their hands to handle other things. If your steps exceed three, others won't want to use it.
Kang Zhengzhong: So Mr. Jiang actually has the capability to solve specific scenarios, it is just that the hardware entry point currently chooses glasses, right?
Jiang Dawei: Our company's dream is "To make everyone a superhero." We hope to output solutions in wearable categories, such as buttons, lapel pins, wristbands, rings. We are not a glasses-making company; we are an AI wearable company.
Kang Zhengzhong: Mr. Jiang is a good entrepreneur and also a good salesperson. Then ask Mr. Hua: everyone mentions AI cannot be separated from privacy. You mentioned that some heavy users are overseas business people, and one advantage of using hardware is that data is on-device. Is the willingness of these overseas business people to pay for on-device privacy encryption enough to support the globalization growth of the translation-recording hardware category?
Hua Kun: Privacy is definitely a concern for many users. But let me give the answer first: I think this market will be relatively small, a Niche market. Users who completely care about privacy will not trust the cloud and require complete localization. Is there a solution now? Like the Gemma 4b model released by Google, it runs well on a 32G Mac mini, and combined with recording equipment can achieve no need for the cloud at all. But the problem is high cost, high threshold, and sacrificing portability. You will find that actually users, in the face of efficiency, can sacrifice a certain amount of security—as can be seen from the OpenClaw case. This is a balance.
In specific scenarios, specific institutions will pay to customize private solutions, but the construction cost is very high. We focus more on "relative security." Solving long meetings still relies on the cloud; short-time instant translation can use local. What we must do in the cloud is ensure storage encryption, while handing the choice to the user. We made a function that only processes in the cloud but does not retain memory. Because the value of large models is remembering who you are—if it doesn't recognize you the next day and you have to explain again, it is indeed hard to choose, so hand the choice to the user.
Kang Zhengzhong: First solve the efficiency problem then consider the privacy problem. Ms. Li, wellness hardware has access thresholds and even legal liability overseas. Taking companion robots as an example, where do you think the commercial explosion point is—B-end (nursing homes reducing costs and increasing efficiency) or C-end families? What is the repurchase logic?
Li Shuting: I think the explosion point for companion robots is definitely at the B-end. The B-end's pain point is the most painful right now—for example, the labor efficiency of caregivers in nursing homes is very expensive, and the overall risk is large. In our scenario, for example, Grandma Zhang wants to drink water, she can reach a caregiver with one tap through a tablet. Because of scenarios such as dialect and cultural differences, the elderly may often easily have conflicts with caregivers. We can provide a companion that understands the elderly and is not intrusive. The C-end is currently not a particularlyrigid demand scenario—it may exist more among high-consumption, high-net-worth人群. The repurchase logic is mainly to cultivate a habit. The sunk cost is also high, so currently the B-end is a relatively large explosion point.
Kang Zhengzhong: Teacher An Shanzi, regarding the current overseas strategy for XR hardware, are you more optimistic about the growth of B-end (enterprise training) or C-end (entertainment and social)?
An Shanzi: This needs to be looked at from both ends, because the hardware forms are completely different. If it is MR headsets (Meta Quest, Pico Neo, Apple Vision Pro this kind), what can truly drive volume is still the C-end entertainment form. This year the MR form will have big changes—weight will drop from the previous 400–600 grams to 100–200-plus grams, the computing unit is moved down, supporting SLAM spatial positioning. The experience improves but the price cannot come down—it may still be over US$10,000, so this year MR will be relatively cold.
The other line is thin and light AI glasses. Prioritize wearing experience, then define tool-type products for specific scenarios. For example, the Even Reality product just mentioned, with weight achieved at 38 grams, no camera or speaker,specifically doing information prompts "head-up display." It sells for US$600–800 overseas and has received relatively good market feedback. Pull wearing and design to the extreme then stack functions, penetrate one scenario thoroughly, and you can get very good market feedback.
Kang Zhengzhong: Due to time, the final open question:假设 today in 2026, we want to All in one AI hardware product, what would you choose? You can choose your own. Mr. Jiang, start.
Jiang Dawei: This year we will definitely still focus on selling these glassesequipped with DigiWindow. Additionally, we have公开 a product called the VOCCI ring, because many people do not want to wear glasses but want to use recording. This ring can record for eight hours. I have had it on all day today, and it can highlight important topics. At night when I turn it off, it outputs meeting minutes for all the day's visits.
Hua Kun: We will still focus on voice. I think there are two very interesting ones. One is the "Always On" scenario that can summarize interactions at any time. Besides general recording, we are thinking about how to make a minimal voice entry point. Because in many occasions you do not want others to hear what you are saying—like Whisper's kind of whispers—hoping AI can understand and free the hands. Voice is a track very worth deep cultivation.
Li Shuting: I would also choose the AI ring. It is无感 to wear, can interact with the human body 24 hours, and has a battery life of up to a week. And it can do personalized DIY customization, connected to the APPdevice to give a personalized desktop. From a girl's perspective, first of all it is a piece of jewelry, and secondly its functions are also very useful.
An Shanzi: The more long-term sexy one is brain-computer interface, but it is too far from consumer grade. I still choose XR. It gives people too much imagination—for example, first solving the replacement of the refractive myopia population. More long-term, it can understand your real world, making you like a creator. Recently we put Fei-Fei Li's world model into a VR headset, and you can generate a world with natural language—this is very cool.