Original title: (Dialogue with Tech King Fangzheng of Tiaoyue | A business gamble for a single-leg robot, leaping from the lab to the market)
Original author: DONGCHA Beating

In June 1984, robotics scholar Marc Raibert and two colleagues published a paper studying a machine with only one leg.

The machine relies solely on a single elastic leg to repeatedly rise and fall on an open, flat surface. Researchers tried to control it from three angles: direction, posture, and jump height. The paper proves that jumping can be systematically controlled, but limited by the technology level of that era, the machine always dragged along hydraulic hoses and data cables, connecting to an off-site power source and computer, and could not operate independently outside the laboratory.

Two years later, Raibert wrote these studies into (Legged Robots That Balance). He started from one leg, then extended the same motion principles to two-legged and four-legged systems. In 1992, he founded Boston Dynamics from the Leg Lab at MIT. Spot didn’t reach full commercial sale until 2020.

From the first single-leg machine jumping up, to a legged robot being bought by a customer—there was a gap of more than thirty years in between.

In August 2026, we saw another one-legged robot. It could achieve stable continuous jumping without relying on any external equipment, and the time from paper publication to a production-ready deployment was under two years. It consists of a carbon-fiber elastic leg and four rotors. When it lands, it compresses the spring. After it’s off the ground, the rotors supplement energy and adjust its posture. If it encounters a ravine it can’t cross normally, it can even fly over it directly. The research team at City University of Hong Kong named it Hopcopter. Zhiyue Technology, based in Hong Kong, chose a more direct Chinese name: “jumping machine.”

Zhiyue Technology originated from the Robotics and Intelligent Systems Lab at City University of Hong Kong. In 2024, papers by Bai Songnan and others about Hopcopter were published in (Science Robotics). Wang Fangzheng then brought in his senior brother Bai Songnan and mentor Professor Pakpong Chirarattananon to turn their research project into a company. The company has already received several million RMB in seed funding exclusively from the Inno Fund, and it is pushing forward with a new round of fundraising. Its first-generation remote-controlled product is aimed at enthusiasts; in the future, it will launch a version that allows secondary development for developers.

Our curiosity is simple: “What is this good for?”

Wang Fangzheng admits that Zhiyue Technology is holding a big hammer and is still searching for the nails that fit it best. He also insists that jumping will become an independent robot form. When investors ask about profit and PMF, he talks about motion principles and future visions. And in the end, he said:

“More importantly, being able to last beyond ten years.”

A single spring

Wang Fangzheng was born in 1999. He earned his master’s degree at Imperial College London in Human and Biological Robotics. After that, he pursued a PhD at City University of Hong Kong. Zhiyue Technology’s earliest technology came from his senior brother, Bai Songnan. Bai Songnan and the team abstracted the process of an animal running—landing, compressing, and then taking off again—into a spring inverted pendulum, and then attached a passive telescoping leg to the underside of a quadcopter.

This route is very much in the spirit of a lab. All the commercial problems Zhiyue Technology later faced started from this spring.

Insight Beating: First, could you introduce yourself, and how Zhiyue Technology started?

Wang Fangzheng: My name is Wang Fangzheng. I graduated with a master’s degree from Imperial College London and studied robotics all along. After graduation, I went to City University of Hong Kong to pursue a PhD with Professor Pakpong Chirarattananon. That’s where I met my senior brother, Dr. Bai Songnan, and I also got to know Hopcopter—what we call the jumping machine.

I think this is a very meaningful research achievement, and there are many possibilities in the future—it may even become a new kind of robot form. Later, I co-founded Zhiyue Technology with my senior brother and mentors, and we’ve kept doing this until now.

Insight Beating: What problem did the lab initially want to solve? And why did it end up making a jumping machine?

Wang Fangzheng: The lab we’re in is a world-class drone lab—we’ve worked on many robot forms. Some drones can roll on the ground like wheels, some can rotate and drift down like seeds, and some can fly motors while suspended by a single rope. We’ve also built drones that can transform, dock, and separate.

A drone can’t escape gravity as long as it stays in the air. Load capacity and endurance are hard to solve at the same time. We thought about using gravity so that once the robot lands, it rebounds. With that, both load capacity and endurance can be improved significantly.

The machine “person” doesn’t necessarily have to look like a person

The bio-inspiration of the jumping machine is embedded in its motion. It didn’t copy the appearance of any particular animal. It only took the moment when humans and animals run—landing, compressing, and pushing off. Wang Fangzheng believes a robot’s form should be determined by tasks, cost, and efficiency.

This judgment also determines how Zhiyue Technology describes its technology. First, it’s a way of moving—responsible for taking the machine to places that wheeled and legged robots aren’t easy to reach. A robotic arm, sensors, and large models can all be things to consider later.

Insight Beating: Why do you think many tasks don’t need to be handed to humanoid robots?

Wang Fangzheng: I’ve always believed robots don’t necessarily have to look like humans. Humanoid robots certainly have their benefits. They can dance, and many real-world environments are designed for human bodies—so humanoids may adapt more easily to those environments.

In a factory, if you need to tighten screws, you can have a humanoid robot reach out and do it, or you can lower a screwdriver from above, tighten it, and then the conveyor sends the product to the next station. I prefer the latter approach, because you also have to consider cost and efficiency.

Many people say the jumping machine isn’t bio-inspired—I have to correct that. It’s highly bio-inspired; it just doesn’t look like the animals we’re familiar with. What it imitates is the moment when a human runs, lands, compresses, and then takes off. We call this motion model a spring inverted pendulum. Bio-inspiration only needs to capture the core feature. When humans build radar, we don’t need to shape the radar like a bat’s ears.

Insight Beating: Compared with quadruped dogs and wheeled/legged robots, which work scenarios are most suitable for jumping machines?

Wang Fangzheng: The two most core advantages of the jumping machine are all-terrain capability and long endurance. Places in the future that need these two capabilities could all become its advantage scenarios.

You can imagine that in future life, there will be many robots. A humanoid robot might be more suitable for at-home use because it can do household chores. In cities, wheeled/legged robots can be used to complete some basic obstacle-crossing tasks. When it comes to outdoors and more complex spaces, there isn’t yet a robot form that can move efficiently. That’s where the jumping machine has a chance to appear.

Insight Beating: Could jumping itself become a barrier?

Wang Fangzheng: We haven’t entered a specific use case yet, so it’s hard for us to describe all of its downsides completely. One limitation we can think of is transporting liquids—jumping will make the liquid slosh.

Humans running itself is also a series of continuous jumps, with the same free-fall experience. What can a person carry while running? The jumping machine will most likely carry similar things.

When it moves, it looks intense, and it can also stop anytime to work in a single place. It doesn’t need to hover continuously. It can support itself on the ground using a landing gear, without consuming extra electricity. There will be solutions later on with legs that can retract as well.

Insight Beating: The jumping robot is still at a very early stage—what metrics should we use to measure its progress?

Wang Fangzheng: Stability is definitely one of them. It needs to stay stable under stronger disturbances and in more complex environments.

After that, the jumping robot will have even more forms. Once we figure out jumping itself, it doesn’t necessarily have to jump on the ground using only one leg. The exact forms can’t be disclosed yet, but you can look forward to them.

Expandability and intelligence are also inevitable directions. In the future, more functions will need to be built together with developers, and it will move forward through the efforts of more people.

Insight Beating: Will you add large models to the jumping machine in the future?

Wang Fangzheng: The endpoint of the jumping machine will definitely include a brain and eyes—so it can move autonomously and also perceive and recognize autonomously. Right now we mainly focus on the lower body, making it jump stably. The sensing, decision-making, and task execution you mentioned all belong to the upper body.

The jumping machine has excellent load-bearing capacity. In the future, we can add things like that to it. We’ll first release a version with a higher level of openness, then move toward autonomous locomotion and autonomous cognition.

The first generation wasn’t called a toy.

For a new robotic form to leave the lab, the first things to solve are mass production, pricing, and durability. Zhiyue Technology chose to start with a remotely controlled jumping machine. It has no autonomous navigation—users control direction, height, and whether to fly or jump using a joystick.

A more realistic question is: after the novelty wears off, will users still be willing to keep using it? Wang Fangzheng put the product’s focus on motion performance, durability against drops, and price. The team even gave up the already-built sci-fi outer shell and switched to a more exposed carbon-fiber structure.

Insight Beating: With such cutting-edge technology, the first step is still a remote-controlled product for enthusiasts. Would the research team feel like it has turned into a remote-control toy?

Wang Fangzheng: First, let me correct one thing—it’s not a little toy. The first generation really is a remote-controlled product. But when you look back at Unitree and DJI, their earliest breakthroughs also relied on remote control.

Unitree’s earliest Laikago didn’t have nearly as many capabilities as today—it was mainly remote-controlled. Back then, it sold for tens of thousands of RMB, and the high price made people think it was very “professional,” which also made it hard to enter the mass market. But that doesn’t mean Unitree didn’t want more people to see and use it.

Our advantage is that the jumping machine can rely on Shenzhen’s mature drone supply chain to produce it, lowering costs. We want to meet people with a price that users are willing to accept, so more people can see and use the jumping machine.

Insight Beating: Can you disclose the current pricing? There are already colleagues in our office who want to buy it.

Wang Fangzheng: It’s not convenient to disclose it yet. We did questionnaire research—those who really want to buy it basically can accept this price, so there’s no need to worry too much.

Insight Beating: Jumping is a pretty intense movement. Will the spring, the body, and the rotors wear out easily? Will users need to replace parts frequently?

Wang Fangzheng: We’ve considered these questions. During internal testing, we played very aggressively. The latest footage was shot in South Africa—we took it to sea rocks and the wild to jump. The machine could move normally and could even jump continuously for a long time.

There was a major shift in our product thinking. At first, we wanted to make a product with a very sci-fi appearance, similar to the concept machine in the promotional video. The machine could already jump and also demonstrate. Later we spent a long time playing with it ourselves and found that people always want to do some intense and exciting operations—falls and control mistakes are hard to completely avoid. The shell is easy to get damaged in these situations, and the distributed weight also affects control stability.

We rethought what mattered most to the first batch of users. First, they cared about jumping itself, and they also cared that even after it falls, they can keep playing with it. In this version, the main material is carbon fiber, putting motion performance first.

Insight Beating: What level of control precision can it achieve right now?

Wang Fangzheng: There’s an indicator that can show its stability. Without giving it any directional control, it can basically perform point jumps within a relatively small range—it won’t jump farther and farther.

This generation doesn’t add extra sensors, so it can’t yet accurately jump how many centimeters, or land on which exact point. The control logic is very simple, similar to remote-controlled cars. Push the joystick forward and it goes forward; push it back and it goes backward—consistent with people’s intuition for using a remote-controlled product.

Insight Beating: Many robotics companies start with B-end customers. Why did you start with the mass market?

Wang Fangzheng: There’s a saying: “If it can reach consumers, don’t do B-end.” B-end usually requires lots of resources, and the reasons behind it are quite complex. For direct-to-consumer, you can validate the product directly with the market—everyone who wants to buy has a chance to buy it.

We seriously considered doing direct-to-B, and some B-end customers did come looking for us. But our team is small, and the technology is still very early. If we started with inspection work right away, the team would have to spend a lot of effort researching the inspection equipment mounted on top. Our advantage lies in jumping motion itself, not in the inspection equipment.

If you start by doing inspection robots right away, the time we spend refining the jumping technology will be reduced, and it’s easy to fall into a dead end. We think this path is better: first make a consumer product, then launch a version with a higher degree of openness so more people can develop together.

Insight Beating: Your first-generation product mainly serves enthusiasts. How do you hope it forms a community?

Wang Fangzheng: The community can be divided into two parts. The first part is enthusiasts—they’ll share creative ways to play. The second part is developers—they’ll add features on top of the machine, such as following, obstacle avoidance, or integrating a large model.

The first-generation product doesn’t have second-development capabilities, and mainly serves enthusiasts. Later there will be a semi-open-source version, mainly aimed at developers.

Looking for nails with a big hammer

Get the technology first, then find the problems—this sequence runs through the company’s products and fundraising. When you look at investors in consumer electronics, they often ask first: who will it sell to? Investors in hard tech are more willing to look first at how well it can move and where that mobility can take it.

Wang Fangzheng summarized this disagreement as “seeing near” versus “seeing far.” This line carries the perspective of entrepreneurs, as well as the reality that an early-stage company needs funding. The company wants to continue doing research, prepare for mass production, and also prepare for the product to enter the market—every one of these things costs money.

Insight Beating: You have the technology first, then look for the problems it can solve. That’s not the typical startup order for many companies.

Wang Fangzheng: That’s exactly the difference between us and most companies in the market. Many companies first spot a pain point, then look for technology that can solve it. But we already have a very powerful big hammer in hand, and then we think about which nails we can hit in the future.

In 2024, humanoid robots just started to become popular. At that time, many humanoid robots walking systems weren’t stable yet—they needed someone to support them, or they could only walk on flat ground. In the lab, we had already been exposed to some breakthroughs that had not yet been made public. A jumping robot can quickly run around campus, cross obstacles, and adapt to different terrains. I think this is a major technical breakthrough in the robotics field, and the possibilities in the future may be broader than those of humanoids.

From day one, the company has been discussing where it can be used. There are many possible directions, but at this stage it’s too early to jump directly into a single use case. What we’re very clear about is that this technology may have the chance to change the world in the future, and we want to push it forward.

Insight Beating: When the use case hasn’t been fully defined yet, why were first-round investors willing to invest in you?

Wang Fangzheng: Investors who are willing to invest in us—this counts as being willing to invest early, invest small, and invest in hard technology. For many pieces of hard tech, it’s hard in the earliest stage to clearly say which specific use case it has. So investors look first at the technology itself.

Compared with the seed round, this round has made huge progress on the product, and our thinking about the path is clearer. Start with enthusiasts, then form a community—so everyone can jointly develop the technology and explore use cases together. We’ve already mapped this path and a few milestones more clearly.

Insight Beating: Would you define Zhiyue Technology as a commercial company, or as a new kind of lab?

Wang Fangzheng: We’re a technology company. On the company side, we handle the transformation of research成果 into products—make the product well—so that the technology can enter everyday life as quickly as possible.

The lab is also a very important part. I’ve stayed in close cooperation with my senior brother and mentors. They’re now teaching at universities. New results from the lab are handed to the company for transformation. Zhiyue Technology can continue to stay close to the most cutting-edge jumping technologies.

If you look only at the company side, the spending on research and commercialization is roughly half and half. On the lab side, it does pure research.

Insight Beating: You just joined the ecosystem of Guaiguai Robotics and also participated in “Gravity DemoDay 2026.” What support do they need most from you right now?

Wang Fangzheng: We’ve just joined Guaiguai Robotics’ ecosystem. Attending that DemoDay was mainly to get to know more investors. The company is working on a second round of fundraising, and at this stage what it needs most is funding.

With the money, we can hire more people and make the technology develop faster. We need to keep investing in research, prepare for mass production of the product, and also start building the market.

Insight Beating: How was your DemoDay experience?

Wang Fangzheng: I’m very grateful to them. When we registered, it seemed like most of the projects were already decided, and on the poster there were only the last two spots. At first I thought I might just be there to fill a slot.

Later we proved our product really is amazing, and I also explained it well. In the end we got third place, and took home a 10,000 RMB prize. I’m quite happy.

Insight Beating: How is the new funding round progressing? What’s the most common question investors ask you?

Wang Fangzheng: Overall, things are going fairly smoothly. When many people come in, their first question is: what can this do, and where are the use cases? If someone asks like that, then in the end, they probably won’t really be interested. They may be thinking long-term about consumer electronics, but they haven’t understood the issue from a hard-tech perspective. Based on our survey data, our first-generation products can sell quite well, which will give the company more “blood.” But what we hope more is that investors will support our future strategy from the perspective of technological progress.

Everyone is interested in the jumping machine, and the debate is huge. I heard that internal debates at some investment institutions are similar to the comment sections on our Douyin (TikTok China) page. One group cares more about how soon they can become profitable, and when they can find PMF. Others genuinely love the development of technology—they believe this technology will definitely be useful later on, and they’re willing to push it forward first.

Insight Beating: How do you view the conflict between these two judgments?

Wang Fangzheng: I think it’s about seeing near and seeing far.

Unitree is a great example. Many investors I’ve talked with—some previously invested in Unitree, and some wanted to but didn’t manage to—found it wasn’t easy to raise money for Unitree before it appeared on the Spring Festival Gala. Before and after it became popular, it was hard for people to clearly articulate its most certain use case—other than university lab experiments and the developer market. The biggest change was how the market viewed it.

Zhiyue Technology is possibly even earlier than Unitree was in its very early stage. When Unitree started, there was already Boston Dynamics ahead of them. We’re a bit like when Boston Dynamics first released those robot-dance videos—back then, people had just started to see something like this. To make the world understand how impressive the jumping machine is and what it could do in the future, the road is still long.

Solid inside the bubble

Robots are one of the hottest startup directions in 2026. Wang Fangzheng admits there’s a bubble in the industry, and some people advise him to follow the heat and tell an even bigger story. Start by getting a batch of B-end intentions, then scale up the fundraising amount. Once the money is in hand, slowly look for direction. This approach didn’t convince him.

He likes Xiaomi and Unitree, and his reasons boil down to the word “solid” (“实在”). His assessment of Musk follows the same line of thought: he admires the ability of someone to build technology, while also keeping judgments about how to make money and an entrepreneur’s responsibility.

Insight Beating: Which technology company do you like the most?

Wang Fangzheng: Xiaomi. First, I really admire Lei Jun. From early startup setbacks to Kingsoft and then Xiaomi, and finally Xiaomi Auto—he’s been striving and realizing bigger personal life value.

Xiaomi is a very solid company. Once it enters an industry, it often reshuffles the market—many pieces of information and prices become more transparent. A tech company should bring tangible benefits to ordinary people. I think Xiaomi has done that.

Insight Beating: Is there a bubble in the robotics industry right now?

Wang Fangzheng: I think there’s a lot out there. Other FAs have come to us and said the story we’re telling isn’t right. They feel we should be like some robotics companies that raise money quickly: get lots of B-end cooperation intentions first, and then tell a big story. They have already got 5 billion RMB in hand, and they can develop however they want later.

For now, we don’t want to do it that way. We recognize a path, then we’ll talk about that path.

An automobile is a very good kind of robot. The roads it travels every day are basically fixed, and buildings and roads don’t suddenly change. With four wheels underneath, the dynamics are relatively simple. Also, every day there are so many cars collecting data—yet as of today, autonomous driving still hasn’t been fully realized.

A humanoid robot’s structure is much more complex. You have to invest in production, get it into homes, and there’s a long road ahead. The jumping machine’s dynamics can be fully analyzed, with lower costs and a simpler structure. We have the opportunity to get the jumping machine into real working scenarios earlier than humanoid robots.

Insight Beating: What do you think of Musk?

Wang Fangzheng: In America’s capitalist environment, he knows exactly how to play the entire set of commercial rules. He has tech ideals, and he also understands how to make money—he’ll use the same playbook as in the coin world, too.

If you only look at how much he helps the tech field and his entrepreneurial achievements, I’m very impressed by him—he really is great. But when evaluating everything about a person as a whole, I hope that in the future someone can do even better than him.

Live on after ten years

On the morning of the interview, Unitree Technology was listed on the Shanghai Stock Exchange’s STAR Market. In Wang Fangzheng’s WeChat Moments, investors posted congratulations one after another, and some even made money from this investment.

Zhiyue Technology is still far from that day. The company is still collecting user feedback for its first-generation products, plans to find KOLs for closed testing, and will go to the United States to exhibit. It just won third place at the Guaiguai Robotics DemoDay as well, and is now preparing for a second round of fundraising and meeting investors. Wang Fangzheng talks about DJI and Unitree ten years from now, and he’s also clear about what the hardest constraints are right now.

Insight Beating: What’s the startup environment like in Hong Kong? What support have you received from the government and schools?

Wang Fangzheng: Hong Kong provides strong support for tech startups. Our first investment came from Inno China Hong Kong. The funding behind it is connected to the ecosystem in Hong Kong that supports startups.

We were also selected for the training program of HK-Shenzhen Innovation and Technology Park. The park is located at the border between Shenzhen and Hong Kong, and it will provide office space and subsidies. At the policy level, technology startups are strongly supported.

Schools are also important. City University of Hong Kong has the HK Tech 300 program, which supports many startups. For students, the first step in entrepreneurship is hard to take, so the school’s support is crucial.

Insight Beating: What are you planning to do next?

Wang Fangzheng: Recently we’ve been mainly doing user research, collecting people’s expectations for the product. We’ll keep refining the productized features so the first generation becomes more usable.

After that, we might find KOLs for closed testing. The product is expected to be released in early 2027 after it finishes CES.

Insight Beating: Ten years from now, what kind of company do you hope Zhiyue Technology becomes?

Wang Fangzheng: If I had to choose a reference, I’d pick Unitree and DJI.

If in the future a very clear use case appears—one where we have absolute advantages—like how aerial drone (for filming) is for drones, we would independently develop the relevant technologies specifically for that use case, and strive to reach an 80% market share. That would be a DJI-style path.

Ten years from now, if the jumping machine still doesn’t have a very large, clear use case, it will take many people working together to explore. We might then be more like Unitree.

More importantly, being able to last beyond ten years.

Insight Beating: What’s your biggest takeaway from Unitree’s listing?

Wang Fangzheng: I see many investors in my Moments wishing congratulations. They may have made money along the way too. Unitree walked a very solid path. Among these robot companies, the one I like most is Unitree—it truly helped drive the development of legged robotics.

Many people invest in Unitree, especially investors in the later rounds, who mainly look at market heat and market share. They may not necessarily believe the technology itself is that cutting-edge, and they may not necessarily believe in the solid, down-to-earth approach to execution. For a technology company, I think that’s a bit laughable.

I’ve met some of these people. If you’d sent them to talk with Wang Xingxing a few years earlier, they probably wouldn’t have invested.

Insight Beating: How big a share of investors do you think investors like this have in the market?

Wang Fangzheng: I’m not saying they’re bad. Investors are responsible to LPs. Of course they’ll invest in projects within their cognitive range—projects they can understand—because that makes it easier to see clear returns. I fully understand.

Investors like that still account for a high share of the market. In RMB funds, the proportion might even be higher.

What Raibert wrote in (Legged Robots That Balance) is how a single leg maintains balance. Back then, no one could directly see the Spot from more than thirty years later on that machine that kept jumping up and down. The work that research left behind was a set of principles about dynamic balance. It later grew two legs, then four legs, and also gave rise to a company.

Wang Fangzheng is taking a more direct path. He already has a machine that can jump. Next, he needs to turn it into a product that people can buy and that people keep using. This process won’t be achieved by just one impressive jump.

Each time the machine lands, the spring compresses. If you want to jump again, you first have to catch your own weight.

Original text link