145: 孙宽 of Hypershell: The Birth of the First “Consumer Exoskeleton”
Summary
Hypershell’s most important product decision was not making exoskeletons cheaper, but first rewriting the category’s mental model from “medical brace” to aspirational outdoor enhancement gear. 孙宽 believes the label “if you wear one, something is wrong with your body” would repel both young people and older adults unwilling to age. The outdoors represent exploration, freedom and the real world, combining strong demand with a natural distribution channel through photographers, climbers and other content creators. The company believes consumer exoskeletons remain at a very early stage; the real inflection point will not be a marketing hit, but the product appearing in everyday settings such as subways and supermarkets until it becomes unremarkable.
The first-generation product started from an extremely hostile physical constraint: every additional kilogram of load raises human metabolic consumption by roughly 2%-3%, so an exoskeleton must first offset its own existence. Hypershell made assistance its top priority, then brought the product down to roughly 2 kilograms and made it foldable; first-generation range reached 17.5 kilometers. The power system delivered several hundred watts of peak power, with the longer-term goal of recovering energy through regenerative braking on descents. The second-generation X Ultra weighs 1.8 kilograms, offers 30 kilometers of range, and improves assistance effectiveness by roughly 20%-30% over the first generation. 孙宽’s product formula is the ratio between positive and negative effects: only if that ratio keeps expanding can an exoskeleton evolve from equipment into a basic capability.
Replacing mode buttons with a single power button was, in substance, a deliberate decision to remove the algorithm team’s fallback. Engineers wanted to retain buttons as a safety net for immature AI; 孙宽 insisted, “No,” forcing the team to make the device automatically recognize walking, running, stairs, slopes and cycling. Users can still select modes in the app today, but every manual selection enters the learning loop. The trade-off resembles the capacitive touchscreen on the iPhone: the point is not to prove that the technology exists, but to let ordinary people use it without understanding it.
What truly determined the company’s survival was not whether the first single-motor architecture could reach mass production, but whether 孙宽 would admit that his most original answer was still not good enough. Gray testing after the crowdfunding campaign produced scores in the 40s: simulation had shown the single motor to be energy-efficient, but it could not solve problems involving joint-axis misalignment, body-size variation and backpack compatibility. The company ultimately rebuilt the product around dual motors, nearly doubling costs and delaying delivery by another year. “The conservatives are always right, but the optimists may find a more correct direction over the longer term.” The cost was 40 months from founding to first delivery; the payoff was avoiding a pile of bad reviews that could have exhausted both the company and the category.
The financing story also reveals the market’s risk and its validation path: when only RMB200,000 remained, payroll was about to stop and a co-founder was mortgaging his home, the only visible signals were crowdfunding data and users repeatedly saying the product could be “life changing.” Hypershell went from a solo startup in 2021, facing sustained investor skepticism, to two rounds totaling $70M, a valuation of several hundred million dollars and cumulative shipments in the tens of thousands. The first roughly 4,000-5,000 units did not begin shipping until August 2024; regular in-stock sales started the following February. Capital validation came from the product entering the real world, not from a hot sector arriving first—which also means today’s financing enthusiasm cannot substitute for continued validation.
Hypershell has established strong first-mover brand recognition, but has not come close to proving the mass market: 孙宽 estimates that the entire industry generated only several hundred million yuan in revenue that year, with roughly 90% coming from Hypershell. An early user said that on a hiking trail in the U.S., roughly 7 out of 10 people could recognize the product as Hype. An official Facebook community of nearly 10,000 members and a user-built community of roughly 4,000 have turned Walmart cart pushers walking 70,000 steps a day, photographers and users with a history of asthma into distribution nodes. Once competitors see PMF, they will inevitably enter; “the only way to win is to move faster than everyone else.” Hypershell’s defense is not simply being first, but the 40 months of accumulated work in safety, ergonomics, algorithms and manufacturing details.
The next stage requires price, organization and product form to cross the chasm at the same time. X Ultra is priced at $1,999, while the current average selling price is just above $1,000. 孙宽 believes the mass-market inflection point may be one-half to one-third of today’s price—roughly RMB3,000-5,000 in China—but emphasizes that current products are not priced on the basis of exceptionally high gross margins; cost reduction will require more in-house development, process substitution and upstream integration. The longer-term goal is not to sell a “shell” forever, but to “make the human body programmable”: Hypershell chooses augmentation over replacement, allowing the physical body to keep pace with free will. That is also part of what 孙宽 means when he says entrepreneurship has helped him “live more solidly.”
Deep dive
1. Hypershell Is Not Trying to Replace People, but to Stop the Body from Limiting Free Will
Hypershell’s “Shell” comes from Ghost in the Shell. 孙宽 understands the shell as a container for the soul and free will: whether the body is biological or prosthetic, the company’s job is to build a “better shell” that expands a person’s world rather than turning the body into the boundary of intent.
The host asked why Hypershell does not simply build humanoid robots. 孙宽 acknowledged that humanoids are “very cool” and that he watches every new video dozens of times, but said the two paths serve different missions: humanoid robots aim to replace people across scenarios, while Hypershell enhances human mobility. “We’re still staying where we are” is not a sign of technological backwardness, but an intentional choice to stand on the human side.
That choice also responds to his longer-term concern: AI and robots may replace large numbers of jobs, and even allow silicon-based species to displace more vulnerable carbon-based species. He does not want to see that future. He wants technology to help people “break free from some of the shackles of being carbon-based,” while people themselves continue to enter reality, explore and create.
2. Phone Retail, Screwdrivers and One Misunderstanding Shaped 孙宽
孙宽 was born in 1991. His parents ran a large mobile-phone store, so the family had abundant phones and computers from an early age. He folded iron casings into four-wheel-drive cars and sawed material from door frames to make helicopters. His first program was written on a phone; when the phone was lost and a new SIM card inserted, it automatically sent the location and the other person’s number to a second phone.
The core of this childhood training was not systematic education, but the sense of achievement that came from “making something.” His grades were extremely uneven: he needed almost no extra effort in science and engineering, while his English and other humanities subjects were poor. His parents were busy with the business and paid little attention to him, intervening only occasionally.
He particularly remembers getting every question right on an English test, only for his mother to assume he had copied and slap him twice. He became even more resistant to English afterward. 孙宽 describes this as “reinforcement learning”: positive and negative feedback in childhood shape a person’s judgment of their own abilities. That is why, when leading teams, he places unusual weight on passion and autonomous choice.
3. Building a Stereolithography 3D Printer as a Freshman Convinced Him That “Impossible” Was Usually a Blind Spot
During university, 孙宽 spent almost all his time in the lab and completed 20-30 innovation projects: controlling water droplets to form patterns in midair, laser-curing 3D printing, and various mechanical and control systems. The school eventually gave him an independent lab beside the gymnasium. Lasers scanned continuously in the dark room; classmates who wandered in described it as “an alien laboratory.”
Around 2010, as a freshman, he developed a stereolithography 3D printer. There were no open-source solutions he could directly reuse, so he had to relearn everything from 3D models to G-code, laser-galvanometer control, layer-by-layer resin curing, base-plate separation and mechanical structures. The first generation of slicing software approached 10,000 lines of code, and the entire system took roughly six months.
The experience left him with two beliefs. First, “technology is magic—there is nothing you can imagine that cannot be built.” Second, he really was good at breaking down complex problems. He later chose to bet on new categories because he believed that so-called impossibilities were merely possibilities people had not yet seen; opportunities to change the world often hide inside the collective default assumption that something cannot be done.
4. Commercializing Water-Drop Subtitles Was His First Collision with the Gap Between Lab Ability and Business
孙宽 tried to start a business while still in school. He and partners from the student science association called their device, which formed patterns from water droplets in midair, “water-drop subtitles.” They imagined installing it in shop windows to attract foot traffic and went from store to store pitching it. But as “extreme introverts,” they often stood outside the door pushing the question of who should go in first back and forth.
Looking back, he believes there was a “huge gap” between a DIY prototype, a reliable product, paid demand and a complete sales loop. The confidence that “nothing is impossible to build” was mixed with the “naive foolishness” of lacking social and commercial experience. The episode forced him to acknowledge that creating a product and building a business are two different capabilities.
5. Rotations and a Modular PC Let Him Run Through an Entire New-Category Cycle Early
After graduating, 孙宽 was invited by a company searching for a similar commercial direction after he shared projects in maker communities. He had not defined his career; he only knew that he liked creating and solving problems. His boss therefore moved him across departments for two or three years: mobile apps, innovation projects, mechanical structures, smart glasses, bicycle rearview mirrors and server backends.
Around 2015-2016, he led the productization of a 3D printer with a team of dozens, facing clear delivery dates, real customers and the mass production of electromechanical systems for the first time. The experience upgraded “I can build it” into “how do I organize dozens of people to build it on schedule?”
He later incubated an x86 Windows modular PC independently. Low-power Intel chips created an opportunity for TV-stick-style hardware, while Windows developers lacked a platform comparable to the Raspberry Pi. The product generated roughly RMB4M-5M in revenue on Kickstarter without spending a cent on advertising, followed by an independent website, community, trade shows and distribution. Annual revenue approached RMB100M.
The project gave him a full run through the cycle from idea and category validation to retail scale-up, competitor entry and market commoditization. It was still not what he most wanted to do. 孙宽 later called the experience a “simulation” of entrepreneurship, but it gave Hypershell its first muscle memory in consumer hardware and global crowdfunding.
6. A Misdiagnosis Turned “Life Is Short” from a Maxim into an Operating Constraint
During the modular-PC project, a misdiagnosis combined with intense work pressure sent 孙宽 into escalating anxiety about death. He went home to rest for a while and realized that although this episode might be false, “the next one might be real.” He no longer wanted to spend his limited time on a direction that was not his genuine first choice.
The seed of the exoskeleton had been planted earlier. His first hardware project in middle school was an EMG sensor for detecting muscle activity, a direct way to control an exoskeleton. He also repeatedly remembered a red cape from childhood dreams—one that could fly anywhere. The core feeling was not power, but freedom from physical limits.
His personal filter became: if he suddenly died while working on something, would he still feel in retrospect that “every moment had been solid and meaningful”? The exoskeleton was an attempt to solve the fragility of the flesh while bringing his childhood red cape, science-fiction imagination and fear of death into one entrepreneurial direction.
7. Two Years of Simulation Proved the Exoskeleton Feasible—and Created the Answer That Later Had to Be Overturned
Around 2018-2019, exoskeletons still looked like industrial equipment: hydraulic or pneumatic systems, huge boxes on the back, and controls dependent on switches and fixed gait patterns. More critically, every additional kilogram of load raised human metabolic consumption by roughly 2%-3%. A 10-kilogram device could add 20%-30% to energy consumption, making the wearer more tired.
孙宽 saw this as a systems problem spanning robotics, exercise science and human engineering, not something that could be solved through maker-style trial and error. He therefore pursued graduate study in exoskeletons and focused on simulation analysis. Traditional human experiments were easily obscured by sensor error and the black box of the human body; validating one prototype could take six months to a year, while a simulation could produce a result in roughly five minutes.
The research first showed that exoskeletons were theoretically highly feasible from an energy perspective. It then found a counterintuitive optimal configuration: one motor assisting both legs. 孙宽 says the team later built “the world’s first single-motor exoskeleton” on that basis. Its key value was reducing weight, cost and complexity simultaneously, making consumerization possible.
But the research also embedded an important lesson: simulation can find the energy optimum, but it will not automatically tell you about comfort, body-size compatibility or joint-axis misalignment. 孙宽 later abandoned the single-motor mass-production plan, demonstrating that theoretical optimality is only one part of the product answer.
8. Changes in Supply and Demand in 2021 Suddenly Raised the Opportunity Cost of Staying in School
孙宽 had originally applied to an innovation product-design program in the U.K. to continue studying design thinking: understanding needs through empathy, then validating solutions through rapid prototyping. The method remains central to Hypershell’s product development, but in early 2021 he decided to leave school early and start the company.
The supply-side signal came from robotics infrastructure. Since 2018-2019, open-source quadruped-robot papers, new actuators and control technologies had gradually matured. The critical robotics modules for exoskeletons could finally become extremely lightweight and more natural to control. He worried about missing the window.
The demand-side signal came from pandemic isolation. Being confined to a room for an extended period made him feel that human happiness depends heavily on connection with the real world. Aging, urbanization and digital immersion were also pushing people toward interaction through their eyes and fingers; as average age rises, average mobility may decline. “Increasing human mobility” was therefore not just an outdoor-equipment opportunity, but a response to a future way of living.
9. Hypershell Started from a Borrowed Desk with Almost No Capital Consensus
When he founded the company in 2021, 孙宽 had no co-founder and no formal team. He borrowed a desk in a friend’s shared office and built rapid prototypes alone for several months. The office expanded from one desk to several, while the team formed gradually through validation.
At the time, global exoskeleton sales totaled only roughly 2,000-3,000 units a year. The products looked more like laboratory or maker projects. China had previously seen several industrial and medical exoskeleton companies, but their development was unimpressive, reinforcing investors’ doubts about demand scale and technological breakthroughs.
Before 奇绩创坛 provided the first financing, he also received a job offer. After a steady stream of negative feedback, he agonized for one or two weeks and rejected the job only one day before the deadline. His girlfriend made the decision for him: “Even if you don’t choose it today, once you catch your breath, you’ll still go start a company.”
孙宽 understands 奇绩’s selection logic this way: it does not first bet on a direction, but looks for people who can see opportunities where everyone else instinctively says no, while genuinely understanding the industry. 奇绩 may not have seen consumer exoskeletons clearly in advance; it was willing to believe that 孙宽 had seen something others had not.
10. The First Co-Founder Asked Only “Prove It”; One Online Simulation Made the Case That Day
During Shanghai’s lockdown in 2022, the financing money could not even be collected at a bank. 孙宽 called a former colleague, the algorithm head at his previous company, and invited him to join. The two had worked together on motion control, AI, image recognition and action-recognition projects. The former colleague gave him one challenge: “Prove it.”
The two were isolating at home. 孙宽 walked him through the simulator online, then they ran a simulation together and proved that an extremely simple structure was indeed effective energetically. The colleague decided to join that night. Around June 2022, as the lockdown gradually lifted, several people rented an office only about one-third the size of the room used for the interview and began turning the theoretical model into hardware.
11. 100-200 Personas and Ad-Click Tests Locked the Entry Point onto the Outdoors, Not Elder Care
During the lockdown, the team brainstormed roughly 100-200 groups of people who might need enhanced mobility, then designed multiple product forms and ran renderings on Facebook and other platforms to compare click-through rates. Outdoor enthusiasts, backpackers, photographers and climbers showed broad demand: walking farther, carrying more lenses and reaching scenery that had previously been out of reach.
Firefighters, professional workers, older adults and parents carrying children also showed clear demand, but their personas were fragmented. Hypershell did not choose the seemingly strongest pain point in the elder market because traditional exoskeletons were mentally coded first as medical braces: “a medical brace means something is wrong with whoever wears it.”
The host pointed out that the need among older people looked more obvious. 孙宽 used hearing aids as an analogy: many parents clearly struggle to communicate but still resist wearing them, because accepting the product feels like admitting they are old. A negative label not only blocks younger people; it also sets a ceiling for the aging market itself.
The outdoor frame of reference represents the real world, exploration and freedom. It is positive, active and aspirational. Even when an older user buys the product, what they share is “I finally climbed the mountain I could climb 10 years ago again,” not that they need elder-care equipment. Hypershell therefore defined the category first, then allowed the technology to generalize into more scenarios.
12. The First Generation Had to Be Genuinely Useful Before Hiding the Controls and Anxiety
孙宽’s first requirement was not technological showmanship but “genuinely useful—ideally genuinely good to use.” The architecture had to follow biomechanics, while the power system needed to deliver several hundred watts of peak power in extreme movements. For reference, a professional athlete can reach roughly 1 kilowatt during an acceleration sprint; an exoskeleton must approach the power level of real movement to create a noticeable effect in climbing or cycling.
Range had to cover a full day, ideally with regenerative braking on descents. A hard hike requires roughly 20 kilometers; the first generation ultimately reached 17.5 kilometers, not fully meeting the target but clearing the threshold. The later X Ultra reached 30 kilometers.
The most important interaction was AI. Older devices required users to tell the machine through multiple buttons when to start, stop and switch terrain. The result felt like “being forced to move.” Hypershell wanted the device to have something like a human cerebellum: understand movement intent first, then provide assistance.
Engineers wanted to retain more buttons as a fallback when the algorithm was inaccurate. 孙宽’s answer was “No.” Keeping only one power button forced the team to make recognition truly usable. He compared it with the iPhone and Nokia: both had touchscreens, but the smoothness of capacitive touch was what made users keep using and recommending the iPhone.
13. Two Kilograms, Foldability and a Lack of Excessive Sci-Fi All Reduce the Product’s Presence
An exoskeleton is a wearable device. 孙宽 required that it look neither like a “high-tech villain” covered in flashing lights nor like a medical brace that makes people feel physically defective. The design had to follow the body’s muscles and natural curves, coexist with yoga clothes, athletic wear and outdoor gear, and support the wearer rather than steal the focus.
The team studied anatomical curves and load paths, repeatedly observing how users combined the device with short sleeves, long sleeves and different equipment. Professional users sometimes preferred a more discreet device. The design question was therefore not simply “does it look cool?” but how users see themselves after putting it on—and how they think others see them.
The product also had to fold into a backpack, becoming a “superpower you carry with you” or a “physical-capacity power bank”: leave home without wearing it, then take it out when tired. Early engineering prototypes weighed 4-5 kilograms even with a single motor. The mass-production version used a system-level optimization of carbon fiber, titanium alloy, aluminum alloy and battery efficiency to bring the weight down to roughly 2 kilograms.
孙宽’s overall principle is the ratio between positive and negative effects. The exoskeleton’s core functions are assistance and joint protection; range, intelligence and portability are not add-on selling points, but ways to eliminate the new burdens created by the device itself. Internally, the product must reach at least 70-80 points before it is ready to ship.
14. When Only RMB200,000 Remained, Crowdfunding Signals Were More Persuasive Than Financial Safety
The period around the first-generation launch was the company’s most dangerous moment. The team had only 4 people, roughly RMB200,000 in cash, and was continuing to spend on crowdfunding acquisition; by the final month, payroll might not be paid. Users kept commenting that if the product worked, it could be “life changing,” creating important positive feedback from both user responses and campaign data.
Because he believed in those signals, a co-founder mortgaged his home. 孙宽 had no home to mortgage and could only keep investing. Both families chose to support them. 孙宽 emphasizes that allowing a founder to commit fully is itself an extremely rare form of capital.
A subsequent financing round eased the immediate pressure, but from the perspective of the time it was still an “unboxing”: no one had truly experienced a consumer exoskeleton. Crowdfunding proved that users would pay for an imagined product; it had not yet proved that the machine could deliver on its promise on a real mountain trail.
15. Gray Testing Forced Hypershell to Abandon Its Most Original Single-Motor Design
After crowdfunding generated several thousand orders, the team took a small batch of Hypershell Go prototypes into real-world outdoor gray testing. Users stepped onto very high rocks, carried heavy packs and made large, highly agile movements. The product worked on metabolism, weight, range and assistance, but actual scores were generally only in the 40s.
The root cause was the architecture. The single motor helped the swinging leg through the supporting leg and was highly effective energetically, but it misaligned the motor axis with the body’s joint axis and created compatibility problems with backpacks, body shapes and movement. Smaller, thinner bodies were especially difficult to secure, making compatibility worse for Asian women. 孙宽 himself is very thin and later became the team’s “half-size” testing benchmark.
The internal debate was intense. The original design was close to mass production and could be delivered relatively quickly; the new design required a second electric-drive system and nearly doubled costs. 孙宽 still insisted on dual motors because users do not care about technical originality; they care whether the product is comfortable and whether they want to keep using it.
The rebuild took not the expected six-month patch, but close to a year of development from scratch. 孙宽 later acknowledged that the conservatives were right about cost and schedule: “The conservatives are always right, but the optimists may find a more correct direction over the longer term.” He preferred to absorb the delay rather than trigger a “wolf is coming” reaction with a product scoring 40 or 50 points.
16. The Hardest Part of Consumerization Was Not the Motor, but Fabrics and Contact Points the Lab Had Never Studied Seriously
Exoskeletons previously had no consumer-industry standards. Ordinary consumers might fail to tighten the device or loosen it while walking; a powered structure could then spring up and injure them. The team spent two or three months developing an algorithm that continuously judged whether the device was actually being worn.
孙宽 believes the hardest part to design was not the seemingly complex robotics, but “every place that touches the human body.” After tens of thousands of repeated loading cycles, tiny pressure points in belts, straps and fabrics become friction, pain and even bleeding. Almost no one had previously studied how to wear an exoskeleton comfortably for 7-8 consecutive hours.
The team looked across industries to bicycle saddles. Long-term load does not necessarily belong on the part with the most flesh; it may be better placed on areas with less soft tissue, near bony connection points. The hollow of the waist therefore became a key support point for the rigid belt. The design looks simple, but rests on repeated human testing.
The process gave “defining the category” a second meaning. It meant not only defining why consumers buy, but also defining the checklist for safety, wearing, durability and comfort. Every detail that is “not a problem” in the lab can become a systemic risk when placed on thousands of ordinary users.
17. Motion-Intent Recognition Is Usable, but Truly Invisible Interaction Still Has a Long Way to Go
Current products can adapt between walking, running, stairs, slopes and cycling. After roughly 2 steps on a bicycle, the device can turn an ordinary bike into something resembling an e-bike. Multiple iterations since launch have steadily reduced complaints about inaccurate intent recognition; 孙宽 describes the current state as “basically usable.”
The app still retains mode selection, not because the one-button principle has been abandoned, but because recognition is not yet completely accurate. Every manual selection provides a feedback signal for the system to learn from again. 孙宽 sees this as part of a reinforcement-learning loop; the ultimate goal remains to reduce human intervention.
Software is updated roughly every 2-3 weeks. For example, cycling mode evolved from “not really following the legs” into a Plus version that automatically adds power uphill or when the user stands on the pedals. Compared with hardware versions that take 1-2 years, OTA lets sold devices gain new capabilities continuously and shortens the feedback-to-fix cycle.
18. Six Months at the Factory and 2 Million Impact Tests Were the Prerequisite for Putting the Machine on People
Once the dual-motor architecture solved the functional problem, reliability became the next low point. 孙宽 spent roughly 6 months at the partner factory, not merely pushing schedules but participating directly in production and failure analysis. Only after small-batch prototypes were converted to mass-production processes did many previously invisible weak points emerge.
An exoskeleton continuously applies substantial force to the body, so any structural fracture can become a safety incident. Hypershell set its extreme-condition impact standard at roughly 2 million cycles, compared with 孙宽’s bicycle-industry reference of roughly 100,000-200,000 cycles. During validation, “almost every component broke at some point,” and each improvement could require waiting 2 weeks to see the next failure.
The team tried more than 10 treatments for individual components and only felt comfortable shipping after nearly 6 months. 孙宽 later summarized the right scale for a new-product launch: “If it doesn’t make you ashamed at all, it was probably launched too late; but if it makes you extremely ashamed, the company is probably finished.” Reliability testing is the time between those two extremes that cannot be skipped.
19. After August 2024 Deliveries, Users Began Taking Over the Definition of the Product
Hypershell shipped its first products in August 2024, then spent roughly 4 months ramping production from dozens of units a month to hundreds, gradually fulfilling roughly 4,000-5,000 crowdfunding and subsequent orders. The second half of 2024 was essentially free of formal external sales until regular sales launched in the “February of this year” referenced by the program.
The company built a Facebook community approaching 10,000 people. In the early delivery period, the first thing 孙宽 did after waking was check new-user feedback. An emergency-response group was created, with cross-functional standups roughly every 2 days to determine whether a problem should be solved through software, hardware or accessories.
The user who impressed him most was a Walmart cart pusher who walked roughly 70,000 steps a day and regularly ranked near the top of the step leaderboard. The exoskeleton significantly reduced the burden of his work. He later became a product ambassador voluntarily, recommending and letting people try it wherever he went. The example showed that although the consumer entry point was outdoor use, the capability could flow into high-intensity labor.
Other users had not retraced familiar routes with their families for 10 years, or had suffered asthma and been unable to climb for a long time, then climbed snow mountains again after putting on the product. Photographers paired a Steadicam on the upper body with the exoskeleton below to move around basketball courts while filming. These “life changing” stories confirmed that the product had begun creating value beyond the original plan.
20. Noise, Lost Assistance and Left-Right Asymmetry Turned Negative Feedback into a Product Roadmap
The first batch also had many problems: abnormal noises, insufficiently intelligent algorithms, range that fell short in certain scenarios, and even sudden losses of assistance while walking that nearly caused users to fall as their force levels changed. 孙宽 did not downplay the issues, saying every negative comment in the community quickly entered the emergency group.
Software problems could sometimes be fixed through OTA in roughly 2 days. Problems that could not be solved purely in software were addressed with replacement accessories or moved into the next hardware generation. Crowdfunding users therefore became not only buyers, but co-developers helping Hypershell build real-world exoskeleton data and a failure database.
The team later displayed left-right leg asymmetry data for some users and allowed the two sides to be adjusted separately. 孙宽 remembers one video: a person walked normally when opening a store, then began limping after closing. After adjustment, some users’ gait returned to normal.
These cases changed the team’s view of the order of expansion. Workers and people with weaker mobility did not necessarily need to wait years for the outdoor segment to mature before entering. Mobility is a relatively universal capability; the market boundary may expand through users’ spontaneous applications faster than through the personas planned in advance by the company.
21. Forty Months to Define the First Generation, Four and a Half Months to Launch X Ultra
Hypershell took roughly 40 months from incorporation to first delivery. 孙宽 acknowledges that some problems could have been tested earlier and found sooner, but believes new hardware with no historical reference must “respect time”: detours delay delivery, but also build reusable judgment, standards and supply-chain capabilities.
The current product line includes X Go, X Pro, X Carbon and X Ultra. The first 3 are different models of the first generation; Ultra is the next generation. Ultra took only 4.5 months from formal project approval to mass-production launch, around the “September or October of this year” referenced by the program.
Ultra’s weight fell to 1.8 kilograms, range reached 30 kilometers—described by 孙宽 as roughly 3 times the first generation—and assistance effectiveness improved by roughly 20%-30%. Without changing the battery, the team squeezed overall efficiency out of every module “bit by bit.” The first-generation range was previously given as 17.5 kilometers, creating a discrepancy with the “3 times” description.
Four and a half months does not mean only 4.5 months of technical accumulation. The large number of pre-research concepts generated during first-generation development became an “ammunition depot” that could be recombined into a product once the project began. The real change was organizational: the team evolved from a 4-5-person “scrappy outfit” holding the core together into an organization capable of system-level optimization.
22. The Founder’s Full-Stack Experience Let Hypershell Add People Quickly Across an Extremely Long Hardware Chain
孙宽 describes consumer hardware as an extremely long chain: research, demand insight, product definition, engineering, supply chain, mass production, global marketing and distribution. Any link can constrain the outcome. The prerequisite for expanding the team was first using a “monkey version” to run through the entire chain, then identifying what capability each module genuinely lacked.
His early work writing apps and backends, building mechanical systems, managing projects, running crowdfunding and developing channels meant he was familiar with most links. When the team was small, he personally handled every module, which made it easier to know what kind of person to hire. 孙宽 says bluntly, “I think I’m relatively versatile.”
Management later added 2 co-founders. One had led marketing at OPPO and later headed a business unit at Nothing, bringing brand and global commercialization experience. The other had spent 7 years founding companies in Boston, repeatedly asking how users see themselves after putting on the product and how others see them, while pushing user research, product methodology and supply-chain upgrades.
23. “Infinite Improvement” Allows a Rough First Version, but Not Stopping Short of the Extreme
The first layer of Hypershell’s values is passion and drive. Before assigning a project, 孙宽 always asks: “Putting objective constraints aside, do you really want to do this?” The company allows internal transfers because people without interest gradually disengage and spread that state to those around them.
The second layer is the pursuit of strong output. If a consumer exoskeleton is not built to an extreme standard, the result is not merely less attractive; it is directly unusable. In interviews, he asks candidates about the hardest and most satisfying work they have done, looking for evidence that they proactively raise their standards.
The third layer includes truth-seeking pragmatism, rational objectivity and long-termism. 孙宽 still meets candidates personally in the final interview round and first asks how they view Hypershell’s category. If someone is just randomly looking for a job, he believes a large company or a hot humanoid-robot company may be a better choice. Hypershell needs people who genuinely believe in human augmentation.
The underlying word in his WeChat signature, “Infinite Improvement,” is iteration. As a child, after learning the principles of pneumatic cylinders, he used one to build a calculator capable of simple addition within 10 minutes after class. He believes that “running it once matters”; the first version can be a monkey version, as long as continuous iteration can make the next version 10 times faster or 10 times better.
24. A Batch of Mislabelled Products Shifted 孙宽 from Talent Density Alone to Talent Plus Process
孙宽 once believed that high talent density, mission, culture and room to create could substitute for process. Innovation teams, he thought, should not be constrained by the layers of checkpoints typical of automakers. But once products entered scaled manufacturing, missing processes amplified into batch-level problems; a small mistake could cost millions of yuan.
The classic lesson was a batch shipped to Europe with incorrect product labels and missing CE certification labels. The entire batch was returned, and expensive cross-border logistics magnified the loss. 孙宽 also went to a relevant European court for the first time to explain what kind of new category the product was. The issue was not complicated, but became a “boomerang” exposing a management gap.
His reflection was not to find someone to blame, but to admit that he was naturally scattered and had previously underemphasized process. “Process is methodology that has been solidified”: record the most efficient methods discovered through repeated experimentation and establish necessary checkpoints at high-risk nodes, such as scanning to verify labels after application.
The company subsequently built digital and information systems and established a product-manager team focused on AI and informatization to identify processes that could be standardized by systems. 孙宽 still encourages trial and error, treating early losses as the organization’s “tuition.” But if a mistake comes from carelessness rather than an unknown problem, he becomes extremely dissatisfied. “Gentleness is my user interface”; behind it remain high performance, strong incentives and bottom-tier elimination.
25. After Growing from 40 to More Than 200 People, Collaboration Became Scarcer than Individual Ability
At the beginning of the year referenced by the program, Hypershell had roughly 40 people, then quickly expanded to more than 200. 孙宽 at one point directly managed nearly 100 people. Many excellent new hires had no leader and could only report under the founder, but he lacked enough time to define clear scopes, challenges and growth paths. Confusion and coordination friction followed.
The team is adjusting its goal-setting mechanisms, developing nearly 20 team leads and cautiously bringing in external managers. 孙宽 believes strong talent has many options: if people are unhappy for 10 minutes, they can find another job. Management is therefore not about making people obey, but allowing ability to fully express itself against clear goals.
Hypershell has partly borrowed from 拓竹: high performance, strong incentives and bottom-tier elimination jointly maintain talent density and make the environment feel fair to people who actually get things done. 孙宽 observes that 拓竹 employees remain very “full” in spirit despite enormous challenges and do not easily consider outside opportunities.
The host added that DJI employees’ description of “comfort” does not mean low pressure. It means being surrounded by people who do the work, with little office politics and evaluations based on output. 孙宽 agrees with this “purity”: making something good earns respect. It is also why he favors candidates who feel “fresh and youthful.”
26. Crossing the Chasm Requires Content to Move Down Layer by Layer, Not Selling Directly to Everyone
Hypershell started with hardcore outdoor users because their needs were strong and because they naturally produce content. Climbers and photographers share the experience of conquering peaks and reaching unfamiliar scenery; the content attracts broader outdoor and travel audiences, then continues influencing people’s everyday mobility, creating a flow of category awareness from professionals to the mass market.
Each layer of breakout requires 3 things to be true simultaneously: the group must have a strong enough pain point, current technology must genuinely meet it, and users must be willing to spread the experience to the next layer. Chasing traffic before product capability crosses the threshold would prematurely consume the credibility of the new category.
The host pointed out that humanoid robots broke into the mainstream rapidly through events such as the Spring Festival Gala and Tesla Optimus, while sending a Roadster into space with a rocket was a similar form of distribution. 孙宽 agrees that exoskeletons also need events. The company will co-create content with users that could become hits, but such turning points are “opportunistic” and cannot be treated as a controllable growth formula.
The real inflection point in his mind is seeing someone naturally wearing an exoskeleton on a Beijing street or subway without needing to explain it. A friend has already seen a “very mech-style old man” wearing the product, carrying shopping bags and leaving a supermarket in the Shanghai subway. This kind of daily penetration is closer to crossing the chasm than one high-volume marketing event.
27. The $1,999 Ultra Is Still Early-Adopter Pricing; RMB3,000-5,000 May Be Needed to Open the Mass Market
X Ultra is priced at $1,999, while existing products average just above $1,000. 孙宽 says the team has run price tests to observe acceptance at different levels, but current consumers are heavily skewed toward early adopters with greater interest in and tolerance for new technology. Their behavior cannot be used to infer the mainstream market.
He estimates that the critical price may be one-half to one-third of today’s level: moving from the price of an iPhone toward that of a midrange phone, or roughly RMB3,000-5,000 in China. That level may become the inflection point for demand elasticity, but achieving it will require time and new engineering capabilities.
Responding to the claim that gross margins are extremely high and the product could cross the chasm simply by cutting prices, 孙宽 explicitly denies it. The product uses expensive materials such as carbon fiber and titanium alloy that users cannot see. The sales team pushes for cost reductions every day, but he believes the current product is already fairly close to cost. Innovative hardware must also retain sustainable gross margins to fund R&D.
Cost-reduction paths include expanding in-house development, improving processes and replacing some carbon fiber with lightweight materials suitable for mass production. The rear chassis uses the same 3N titanium-alloy process as Apple Watch Ultra. Hypershell has also invested in a related manufacturer, using deeper supply-chain integration to improve production and cost rather than simply forcing prices down.
28. 拓竹 Shows That Complex Hardware Needs a “Usable” Product and a Scalable Content Flywheel to Cross the Chasm
孙宽 describes early 3D-printer users as both users and developers, mostly “playing” with the products. 拓竹’s first inflection point was making a complex device feel as easy to use as an iPhone: automating many operations so ordinary people only needed to press one button. The product then moved from a development tool to a consumer capability.
The second inflection point was the content community. 3D modeling is harder than printing itself, so 拓竹 turned MakerWorld into something like an App Store where creators upload models and ordinary users can use them directly. A good product brings users; users attract models; models drive more use and consumables. The flywheel begins turning.
孙宽 imagines the end state not as “a better 3D printer,” but as a “toy transporter” or even a spatial portal for objects: pen holders, storage boxes, tools and children’s toys produced instantly, cheaply and personally. For a child, the gift from a parent is not just a toy, but creativity and “the memories of an entire childhood.”
29. Exoskeletons Are More Like Cars than Standalone Gear; Basic Mobility Could Create New Ways of Living
The host asked whether exoskeletons could develop an ecosystem like a model community. 孙宽 believes they are more like cars: they provide the underlying capability of mobility. If a pair of legs had “infinite strength and an infinitely long health bar,” travel routes, sports, entertainment and everyday activities could all be redesigned.
He compares it with a single-lens reflex camera. Before buying one, you may have no strong reason to go out; after buying it, seeing the device makes you want to shoot. If an exoskeleton becomes a stable capability, it could drive new mountain routes, activities and communities. Only after cars became widespread did road trips, roadside check-ins and related cultures emerge. The more fundamental the capability, the more downstream possibilities it creates.
The current ratio of positive to negative effects has only “just crossed the balance point,” so the product is better suited to situations such as climbing a volcano or spending an afternoon in a museum—scenarios with clear physical demands—than going downstairs for coffee. Further everyday adoption requires a smaller, more invisible and more effortless product while preserving sufficiently strong assistance.
Beyond consumer use, Hypershell donated devices to post-disaster rescue operations in Hong Kong. 孙宽 relayed official thanks saying that rescuers had to search for survivors and small animals across roughly 30 floors, going up and down dozens of times a day; the exoskeleton significantly conserved their physical capacity. Such extreme tasks show the direct value of mobility in life-and-death scenarios.
30. First-Mover Status Created Brand Recognition, but 孙宽 Says the Current Market Is Too Small for a Zero-Sum War
Traditional medical and industrial exoskeleton companies are moving toward consumer products. Companies such as 傲鲨 have launched consumer models, while Haier released 2 exoskeleton products in the year referenced by the program. 孙宽 also mentioned that more large players may be operating below the waterline and that Hypershell has previously received acquisition offers.
Hypershell’s first advantage is timing. An early crowdfunding user said that on a U.S. hiking trail, roughly 7 out of 10 people recognized the product as Hype. If a consumer’s first exposure to an exoskeleton comes from Hypershell, the brand may gradually become synonymous with the category.
Product experience also drives organic distribution. In addition to the official Facebook community of nearly 10,000 people, there is a user-built community of roughly 4,000 with no direct relationship to the company. 孙宽 believes later entrants may quickly produce products with similar appearances, but may not have gone through the safety, algorithms, soft materials and mass-production details accumulated over 40 months.
He estimates that total industry revenue that year was still only several hundred million yuan, with perhaps 90% at Hypershell—far from a mature competitive structure. One line from the 奇绩 period remains in his mind: “Once you have PMF, the entire market will see it, everyone will enter, and the only way to win is to move faster than everyone else.” The larger enemy today remains inadequate category experience, not any single competitor.
31. The End State Is Not One Exoskeleton, but a “Programmable Body” and a New Division of Labor with Free Will
Hypershell’s lab is also experimenting with “very strange” prototypes such as upper-body exoskeletons to help photographers support a Steadicam. Hardware versions may continue to iterate every 1-2 years, but 孙宽 is unsure whether the mainstream form in 5 or 10 years will still be today’s U-shaped shell. What he believes is that physical-augmentation devices will gradually enter everyday scenarios that everyone considers.
He defines the exoskeleton from first principles as “making the human body programmable.” The device is an input-output interface between people and the physical world. It can enhance strength or simulate experiences; today it can already simulate the feeling of running underwater, and in the future it may be possible to “download a stronger pair of legs” as if downloading software.
The host asked which path matters more: uploading consciousness, becoming prosthetic or retaining the biological body. 孙宽’s answer is not allegiance to any particular vessel, but “mind over body.” Whether the body is prosthetic or supported by an external device is not the core issue; what matters is whether the soul and will are free, and whether the body can keep pace with the imagination. As for whether he would want his consciousness to live forever, he admits, “I can’t answer that today,” revealing only that he and his girlfriend often argue over whether to upload it.
Entrepreneurship did not simply make him happier, but it made him “live more solidly.” He used to wake suddenly in the middle of the night, realizing that one day the self would disappear. That happens far less often now because he is doing everything he can to create the value he wants to create. “That is what it feels like to be living right,” and it is the deeper source of Hypershell’s long-termism beneath financing, valuation and short-term trends.