Huazhe Xu: Embodiment Just Starting Its Race, People Who Won't Wait for IPO | North Slope Project

For the second installment of North Slope, we sat down with Huazhe Xu, founder of PoliTech and assistant professor at Tsinghua University, for an in-depth conversation.

Over the past two years, embodied intelligence has become one of the hottest sectors in tech, with the entire industry caught up in an "acceleration narrative." But from Xu's perspective, most of today's progress has only completed the "first kilometer" on the path to true generality. Many fundamental problems remain unaddressed: where does large-scale data come from, how do models generalize, how are long-horizon tasks executed stably, and how can the complex physical interactions of the human world be genuinely understood?

In a sense, embodied intelligence isn't a sprint-style technological explosion but more like an ultra-long-cycle marathon requiring continuous iteration. Rather than asking "when do we win," Xu cares more about when he can start doing what he wants to do.

"The most expensive privilege is doing what you truly want to do."

What truly draws him has never been arrived-at results, but things not yet formed — the process of constantly venturing into the unknown and pushing boundaries. When the conversation turned to music, he mentioned his love for Beethoven's Eroica symphony, that sense of grabbing a spear and charging into sustained battle.

In some ways, he is that kind of person himself.

Always tinkering, always fighting.

If you have any thoughts you'd like to share, feel free to leave a comment below.

The original podcast transcript was approximately 30,000 words; this edited version is roughly 4,600 words.

Enjoy

Oasis Capital: What's the biggest difference between this startup and what you imagined?

Huazhe Xu: Embodied intelligence has been developing for two or three years now. I didn't expect the market to still be this hot. When I decided to go independent, I was mentally prepared for the market's attention to embodied intelligence to have already passed. The market isn't something I can control — if that had been the case, I planned to start very small. But it turns out the market still believes new companies can accomplish something this year.

Oasis Capital: So the market being this good actually exceeded your expectations. If you had assumed the market wouldn't be particularly strong, why were you still willing to do it?

Huazhe Xu: Building useful robots is our mission. Since we're going to do this anyway, whether the market is good or not isn't particularly important or fundamental. Because the ultimate outcome has nothing to do with initial funding scale — you just plan according to however much you can raise.

Oasis Capital: Though many people probably don't understand your decision.

When the market was skeptical about embodied intelligence, you joined and co-founded a company. When that company might have gone public this year, you chose to start fresh and found your own. By normal logic, you'd wait for the IPO and some returns before leaving, and besides, many people believe embodied intelligence opportunities currently belong only to large companies. With all this external noise, how did you make your decision?

Huazhe Xu: I returned to China in 2022, having just moved from one university to another — Tsinghua. My research group was just getting started, so partnering to build a company seemed like a better choice, and good experience.

But by late 2025 or early 2026, I came to see time cost as the most expensive thing. Even if the IPO would bring great returns, the most valuable privilege is doing what you want to do. So starting a year earlier was completely worth it. For me, financial return just means not starving. My health, having enough to eat and wear — what matters is making things happen.

Oasis Capital: So for you, financial return isn't important; what matters is when you get things done.

Huazhe Xu: I see financial return as a reward for getting things done. Doing something you deeply believe in and that has meaning for society matters more.

Oasis Capital: Because you were an important participant and witness in the previous phase of China's embodied intelligence development. Why do you think there's still opportunity now?

Huazhe Xu: Large model development is like a marathon — if the leaders have already run 15 or 20 kilometers with abundant training data, it's very hard for latecomers to join.Embodied intelligence is also a marathon in a sense, but there are two heavy stones in the way: first, the hardware doesn't work well; second, there's little data. Everyone is pushing these stones forward together, and the pace is very slow.

So embodied intelligence may have started at the same time as large models, but it's probably only reached the 1-kilometer mark. From a technical or industry development perspective, there's still huge opportunity today — the stones are only now showing signs of being movable. In a marathon, if everyone only advanced 1 kilometer before, and you consider yourself a good runner, starting today you should still be able to catch up.

Oasis Capital: Meaning that despite running for a while, the companies that started earlier in embodied intelligence haven't actually gotten very far.

Huazhe Xu: This is an industry-wide problem. Currently, with only about 30,000 hours of data accumulated, no matter how hard you run, you can't do much with it.

When industry pioneers become truly insurmountable, it will be when there's enough data and the systems and models are good enough. That requires massive resources and going through countless pitfalls to understand. Right now, the data isn't abundant, so the industry is stumbling through these pitfalls together.

Oasis Capital: Since you have extensive exchanges with researchers in the United States, how big do you think the China-US gap is in terms of overall model progress?

Huazhe Xu: There's definitely still some gap now. But that's because people haven't particularly focused on this area. If substantial time is spent on models and capability improvement, I don't think the gap would be large. Second, regarding data, it's not the case that the US has richer data and we don't. Finally, it depends on whether the original intention is to solve the general model problem, and whether there's sufficient technical conviction that it can be done.

In my view, the world shouldn't be a case of "oranges are sweet in the south but bitter in the north." For example, I did my undergraduate at Tsinghua and PhD at UC Berkeley, and became a decent researcher; it should be that doing undergraduate at Tsinghua and PhD at Tsinghua, UC Berkeley, or Stanford are all perfectly fine. So I feel domestic researchers are lacking some confidence.

Oasis Capital: Recognizing that the foreign moon isn't rounder.

Huazhe Xu: This is especially important. I think you first need this mindset, and it's the same at the company level.First, don't think about who is China's most PI-like company, or who is closest to PI. The mindset should be: compared to PI, who is the company that can pursue AGI.

With this mindset, you might do things right. Otherwise you're always the one lagging behind.

Oasis Capital: That comes back to cultural confidence, confidence in the path.

Huazhe Xu: This matters a lot. I like to openly share my less impressive parts or weaknesses online, or with my students, because people generally see me as having taken a more glamorous path, but I want them to see "he's actually like this." Then students realize "the foreign moon isn't rounder, this guy is just like this." And they gain confidence.

Oasis Capital: This is actually a huge change we see in the new generation of entrepreneurs and researchers — they see themselves as global innovators from the start, not copying or following overseas, but innovating together. You have your path, I have mine, and who wins in the end is still unknown.

This shift in mentality is itself a structural transformation.

Huazhe Xu: As a researcher, younger is definitely better — the strongest people are in their fifth year of PhD, roughly born around 2000.

Oasis Capital: At a previous Tsinghua researcher gathering, we invited Gao Yang and realized he was already considered older; most attendees were PhD students born between 2000 and 2004. Everyone had strong agency, their questions were very sharp — this is probably a very appealing characteristic of the new generation of Chinese researchers or entrepreneurs.

Huazhe Xu: Everyone has changed, and I think this is excellent. When I was in school, I also didn't want to feel that the professor was necessarily right, or that I had to follow them academically. Gao Yang and I were in the same research group at UC Berkeley, and Alexei Efros was a professor from the neighboring group who gave me particularly good positive feedback. At UC Berkeley's AI group meetings, this professor had poor eyesight and especially liked sitting right under the screen to see. That was a shock to me: a professor who looked forty or fifty years old, sitting on the floor to talk with everyone. And when you presented, he would directly say: "This is wrong," point out what was incorrect, and start debating, while also being very open to people suddenly jumping in on his mistakes.

This atmosphere is now relatively common among this generation of domestic researchers.When I discuss with my students, and they point out my mistakes or where my level isn't sufficient and want to explain things to me, there's almost no psychological pressure — this is excellent. Everyone's potential can be fully unlocked, and only then can you find the best, most correct things.

Oasis Capital: What do you think is the reason?

Huazhe Xu: What people care about now is who can make real, good things. Going through lots of formalities actually has no meaning — everyone can see through it.

Also, knowledge used to come from books and teachers; now it comes from large models, so often you no longer need to seek out teachers. The meaning of teachers is more as peer researchers, just with a difference in age.

Oasis Capital: Can you accept becoming an "old researcher"?

Huazhe Xu: Being old has its benefits — knowing enough old things, accumulated insights, experience-based taste, intuition — these are still very helpful.

Oasis Capital: Actually you're not "old" at all in absolute terms — you're born in 1993, just over 30. AI is indeed in a major historical shift, giving young entrepreneurs and researchers more opportunities. And what you mentioned about learning from large models,this is actually a manifestation of shifting authority — before, authority was the teacher; now authority is AI, "all are equal below AI."

Oasis Capital: We've talked about "shape" many times. In your worldview, what does this mean?

Huazhe Xu: Every institution and individual can grow outward in different dimensions. In a high-dimensional space, you'll find some places bulging very high, some places caving in, some places roughly spherical — everyone has their own shape.

In a university, I feel that if students bulge in certain areas, they should bulge bigger; the caved-in parts can be let go, see your ultimate growth potential. But companies are different — say someone wants to do neuro-symbolic AI, which might be completely different from current business strategy. In school I'd encourage continued growth, but can you still do this at a company?

It's not just at the technical level. At the company culture level, OpenAI is a great example — it's a company with extremely clear shape, and when hiring, it only looks at shape. They hire high schoolers, people without college degrees, people whose research credentials might look poor — but after assessment, if the level is high, they can be hired. However, back when people didn't believe in scaling, even if you were a big name from MIT but believed in small-and-beautiful, believed some theorem could solve everything, didn't really believe in scaling — OpenAI would absolutely reject you, your shape was wrong.

Oasis Capital: Actually when you think and use words, it's a bit like morphology — you mention seeds, shapes, growth. If we abstract further, how did your current way of seeing the world first form?

Huazhe Xu: When I came in, I saw the words "nourishing vitality" outside. When I describe the world, especially people or organizations, I prefer using vital, organic vocabulary, not tools or data.

I think people and organizations are very organic. In my growth experience, I've seen many "stat monsters" from good schools — strong in IQ, or very powerful at some stage — but observing further, the development outcome has very low correlation with their stats at the time, and very high correlation with their decisions. In this process, many who were particularly competitive, with excellent grades, later seemed to suddenly lose their "competitive" mentality, becoming very content with stable, comfortable living. Faced with situations I can't explain, I tend to use shape to describe them.These people didn't choose to keep moving forward in the "tinkering" dimension; they were stat monsters in grades because that didn't require tinkering — it was standards set by others.

So when I look at people, I look at their hearts, which direction they choose to grow. Including when I recruit students or hire for the company, I tend to look at this part.

Oasis Capital: The biggest challenge many people face in the AI era is never finding their own reward function. As you said, many people keep competing, then suddenly realize they don't have to compete, or that outside school there are no rewards, and their whole person collapses. From this perspective, how can a college student today, accustomed to school's reward function, find their own?

Huazhe Xu: That thing inside everyone naturally exists; it's just whether you discover it.

If external signals are too strong, those tiny inner voices become noise instead — this shouldn't happen.**You should capture those small things, make them signal, and push external signals back.**I think this will happen in the new generation's education; Tsinghua is doing this, many places are trying. I'll prepare good signals, but if you have better choices of your own, don't choose what I prepared; only choose what I prepared if you don't know what you want. This can actually give young people better guidance. Without strong external interference, inner things will quickly emerge through continuous trying.

Oasis Capital: This is very much like our opening words from back then: when all the noise fades, the signal naturally appears.

Actually when you throw away others' imposed ideas and work, what you truly love reveals itself. In psychology there's a state called "introjection" — roughly meaning others' expectations for you have been internalized as your own expectations. So you find that some people, even without supervision, are driven by others' expectations that have become enormous self-demands; they've internalized those external things.

Huazhe Xu: Equivalent to external signals affecting your own signals. If it solidly affects you for a lifetime, and on the day you die you don't suddenly slap your forehead and say: "How did my life get influenced by you?" Haha, as long as it doesn't reach that point it's OK. It's not like the signal you're born with can't change — growth environment and other external factors are also important variables.

Oasis Capital: What book has influenced you greatly?

Huazhe Xu: Steppenwolf.

Oasis Capital: What was the insight? Or the biggest impact?

Huazhe Xu: I've actually quoted a small passage from it repeatedly. I don't remember the complete original text anymore, but the gist is: many people come to this world seeking to knit sweaters in beautiful houses, turn on the radio to listen to entertaining programs. And some others want to do heroic things that don't fit market tastes, and everyone calls them fools.

This passage well describes part of what I pursue. I'm someone prone to making strange choices, like why I chose to leave when the company was about to go public, as we discussed earlier.

Oasis Capital: Decisions that seem irrational — that means there must be other things in your reward function.

Huazhe Xu: Yes, reading Steppenwolf, I felt his description hit me. Hesse is truly remarkable — others might only have feelings, but he could write them out, like resonance across time and space.

Oasis Capital: Then when you pass away, how would you want society to evaluate you?

Huazhe Xu: I haven't thought too much about this. I hope to be someone beneficial to society; people don't need to mention me.

But if mentioned, say this person did some positive small things; if a few more sentences are allowed after that, then "this guy's tinkering heart, from beginning to end, was always sizzling."

Oasis Capital: So you want people to describe you as someone who tinkers.

Huazhe Xu: Yes, fighting without end.