GeekPark

7 Things We Learned From 1,000+ Conversations With AI Founders

Fri, 31 Jul 2026 17:34:25 GMT

We've spent three years talking to the people actually building AI companies: founders, researchers, investors, and product teams across China, SF, Tokyo, and London.

Not panel discussions. Not keynote fluff. Real conversations. The kind where people say what they actually think because there's no audience.

Here's what keeps coming up.

1. AI made individuals 10x more productive. It made organizations exactly 0x better.

This is the tension we hear most often and nobody has solved it.

Employees are writing faster, researching faster, coding faster. But the org chart? The approval chains? The way decisions get made? All of that is frozen in 2015.

One COO told us: "I have 2026 workers trapped inside a 2015 bureaucracy. The friction is killing us."

The next big unlock in enterprise AI isn't a better model. It's redesigning how institutions actually work when machines are participants, not just tools.

This is an organizational design problem disguised as a technology problem. The companies that figure it out first will have an advantage that compounds for a decade.

2. The model got commoditized. Verticalization became the only real play.

A year ago, you could still argue for building horizontally — broad platform, lots of use cases, grow into it.

That window closed.

Every durable AI business we've watched being built did the opposite: they picked one workflow, went uncomfortably deep, and made their product irreplaceable for one specific person with one specific problem.

When intelligence itself is cheap and everywhere, the moat isn't capability. It's how precisely you've embedded yourself into something people do every single day.

The race to the bottom on model quality is over. The race to the top on workflow depth is just starting.

3. The "agent" conversation went from speculative to operational — fast.

Two years ago people debated whether agents were real. That debate is over.

The questions now are all operational:

How do you build an agent that runs reliably for hours without human intervention? How does it recover when something breaks? How does it actually improve over time instead of doing the same thing on repeat?

The mental model that clicked for the most teams: a good agent isn't a chatbot with more features. It's a new hire that owns a recurring process and gets better with every cycle.

Delegation, not generation. That's the real opportunity.

4. The best AI products got quieter — and that's how they won.

Early AI products had an instinct to fill space. Long responses, detailed explanations, constant output. The model was showing off. Product teams didn't know when to intervene.

The products gaining real traction now are radically more restrained.

One founder told us: "We removed 60% of our AI's output. Retention doubled."

There's a fundamental difference between a tool you use and a product that's integrated into how you work. The latter knows when you need it — before you ask. That's a design problem, not a model problem, and the teams who've cracked it are building something with completely different retention curves.

5. Memory is quietly becoming the most powerful moat in AI.

The default assumption in software has always been: context resets. You close the app, it forgets you.

That assumption is breaking down.

Agents that persist across sessions, accumulate preferences, remember past decisions, and build a model of how you work — that's a fundamentally different product from a smart chat interface.

Here's the compounding effect: once your software remembers you, the switching cost grows every single day. Day 1, it's easy to leave. Day 90, it knows your workflow better than your coworkers do.

The companies thinking about memory architecture right now aren't just building better products. They're building something closer to an ongoing relationship.

6. Going global is easy now. Making money globally is the actual hard part.

Generating international interest for a well-built AI product is almost trivially easy now. Distribution is nearly solved.

But converting that interest into reliable, durable revenue across different markets? That's where every team hits a wall.

The founders scaling internationally aren't thinking about positioning anymore. They're debugging payment infrastructure in markets where their revenue model doesn't translate, creator monetization structures that vary wildly by region, and compliance requirements nobody warned them about.

The moat is in the operational plumbing, not the product itself. The best product in the world means nothing if you can't collect payment for it in Southeast Asia.

7. China's AI story shifted, and the new question is more interesting.

Three years ago, the conversation was all benchmarks. How close are Chinese models to GPT? What's the gap? How fast is it closing?

That framing is mostly gone.

The question now isn't "can China catch up?" — it's "which layers of the stack will China own?"

The model layer is converging globally. But the application layer, the distribution layer, the vertical integrations, those are diverging fast along regional lines. Chinese AI companies aren't trying to build a better GPT anymore. They're building the ecosystems around intelligence that are specific to how business works in their markets. @natolambert 's visit to china also mentioned this.

This is a more interesting story than "catch-up." It's a story about strategic divergence — and it has implications for every AI company thinking about where to compete globally.

The thread running through all of this.

AI stopped being a technology story. It became an organizational design story, a product design story, and a go-to-market story - all at once. The companies winning aren't the ones with the smartest models. They're the ones who figured out how to turn intelligence into something people rely on by default.

That's a much harder problem. And a much bigger opportunity.

Pipes, billing, and commerce graphs: How China is playing a different AI game

Fri, 31 Jul 2026 17:29:35 GMT

This article was published on 21st, May, at @GeekParkHQ.

This was not a normal week in AI, even by the standards of a field that stopped having normal weeks sometime in 2023.

Three things happened that, read separately, look like routine product news. Google announced updates at I/O. Anthropic made an acquisition. Alibaba reported quarterly numbers with the usual superlatives. Read together, they suggest something about where the competition is actually headed — it's not where most of the coverage landed. And China is playing the game differently.

The Google story got framed as a model story. It isn't, really.

Yes, Gemini 3.5 Flash outperforms the previous flagship on coding and multimodal benchmarks at a fraction of the cost. That's significant for developers doing the math on inference spend. But the more revealing number from I/O wasn't on any benchmark slide. It was this: planning-related queries on AI Mode are growing 80% faster than AI Mode queries overall, over the past six months.

That's a behavioral signal. Users aren't just asking questions. A meaningful and growing subset is using AI to organize decisions, set up tasks, think through logistics. Whether that trend continues is an open question. But if it does, it changes what the product actually is — from an answer engine to something closer to a working layer.

At the same time, Google shipped Managed Agents in the Gemini API: isolated Linux sandboxes, versionable behavior files, a runtime that developers can hand off to Google's infrastructure instead of managing themselves. It's a quiet move. Combined with the consumer-side numbers, it puts Google in an unusual position — trying to own both where users start and where developers build. That's a harder thing to replicate than a better model.

Anthropic's acquisition of Stainless didn't get much attention, which is probably a function of how unglamorous the product is.

Stainless generates SDKs, CLIs, and MCP servers. It has, by Anthropic's own account, powered every official Anthropic SDK since the early days of the API. The acquisition brings that capability in-house.

The reason this matters is less about any specific feature and more about what it signals. As agents become more central to how people use AI, the connectors between agents and external tools — the generated SDKs, the model-friendly interfaces, the way developers actually operationalize API access — become part of the product in a way they weren't before. Anthropic is making a bet that owning that layer is worth doing. It's a reasonable bet.

The China angle this week is harder to read cleanly, which is partly why it tends to get flattened in English-language coverage.

@AlibabaGroup's quarterly numbers were strong. Cloud Intelligence Group external revenue up 40% year over year. AI-related product revenue growing at triple digits for the eleventh consecutive quarter, now 30% of Cloud external revenue. The customer base for Model Studio grew eightfold year over year. These are commercialization numbers from a company with existing scale in cloud and enterprise — a different starting position than a standalone AI lab.More structurally interesting: Alibaba has connected @Alibaba_Qwen directly to Taobao and Tmall's product catalog, which spans more than 4 billion items. Agent capabilities now cover order management, logistics, and after-sales service. The practical effect is that the starting point of a purchase can now be a conversation that has access to the full transaction graph — something Western AI products are still mostly trying to build toward through partnerships.

Then there's a development that deserves more attention than it's received.

China Telecom has begun trial token plans starting at 9.9 yuan for 10 million tokens. China Mobile has launched a universal token service in Shanghai, where users access multiple AI platforms under a single account and pay through their phone bill. It's easy to read this as a pricing story. It might be more than that. When AI usage gets billed through carrier infrastructure rather than through cloud consoles or direct subscriptions, it changes who controls the default relationship between users and AI services. Carrier billing has historically been a powerful distribution channel in markets where it took hold. Whether that dynamic applies here is genuinely unclear. But the structural logic is worth watching.

@Baidu_Inc introduced a new reported metric this month: DAA, or daily active agents. The number itself should be treated cautiously — the methodology hasn't been disclosed and the figures are difficult to verify independently. But the framing is worth noting.

Token counts measure inference consumed. DAA, as Baidu defines it, measures tasks completed. If that framing spreads, it would shift how the industry talks about AI productivity — away from input volume and toward delivered output. That's a different scorecard, and on it, the current rankings look less settled.

The through-line connecting these stories is less about capability than about position.

The model race produced clear winners and is still ongoing. But a parallel competition has opened up around something harder to close the gap on: who sits between users and completed tasks, and how deeply that position is embedded in existing infrastructure, billing relationships, and developer defaults.

In the US, that competition runs through search, cloud platforms, office suites, and developer ecosystems. In China, it runs through commerce platforms, carrier billing, enterprise workflow software, and tightly integrated service graphs. The companies best positioned in each market are not necessarily the ones with the most advanced models.

Whether that changes the outcome of the broader competition is an open question. It does suggest that reading AI market results through model benchmarks alone is going to miss more and more of the actual story.

Wearing the Future: Hypershell and the New Shape of Chinese Hardware

Fri, 31 Jul 2026 16:09:08 GMT

Walk into some popular mountain parks in China today and you'll see something that didn't exist three years ago: people renting exoskeletons at the base of the trail, the way they used to rent walking sticks. The device weighs about two kilograms, has one button, costs about the same as a decent pair of hiking boots to buy outright. Most of the people renting it have no idea who made it. @HypershellTech made it

And Sun Kuan, its founder, has a specific way of explaining why it exists: "Tools determine how large a person's imagination can be." He said this at @GeekParkHQ's Innovation Festival 2026 (GeekPark IF2026) last December, laying out his founding logic publicly for the first time. The product at the mountain park kiosk is the physical answer to that sentence.

The $120M across its Series B and the new X Series launch in May 2026 are milestones in a longer process. The more interesting question is what got the category here - and what this company reveals about a generation of Chinese hardware founders building in ways that look nothing like what came before.

What the product actually does now

The new X Series settles the question of whether consumer exoskeletons are a real product category. They are.

The flagship X Ultra S: 1000W peak motor output, 22 N·m torque, 30km range per charge, 97.5% gait synchronization across varied terrain, 0.31-second response time - 64.5% faster than the previous generation, TÜV Rheinland verified. Weight 2.5kg. One button.

That last detail is the output of HyperIntuition: Hypershell's self-developed end-to-end AI motion control system, and it signals a genuine architectural shift. Previous consumer exoskeletons ran on rule-based gait models: fixed patterns that couldn't adapt to uneven ground, stairs, or the variation between how different people actually move.

HyperIntuition learns. Instead of working in a separate sequence of steps, HyperIntuition works as one continuous system, processing raw data through the neural network, directly outputting the motor torque required for each joint.

It understands your movement as it happens and provides support instantly. Sun calls it the "external cerebellum" - the part of the system that handles automatic coordination so the conscious mind doesn't have to. "Just like the iPhone used the capacitive screen to achieve single-button interaction," he said at IF 2026, "AI (AI agent) lets the exoskeleton achieve the most simplified interaction threshold - users don't need to do anything other than turning it on and off."

Sun Kuan at the stage of IF2026

The 0.31-second response time and 97.5% synchronization aren't spec-sheet decoration. In a wearable system that assists physical movement, latency and synchronization have consequences in a way they don't in software. A chatbot that gives a bad answer is annoying. An exoskeleton that mistimes assistance on a staircase is a different kind of problem. Third-party verification on these numbers, and a 64.5% improvement over the prior generation, suggests real engineering progress rather than incremental polish.

Where it gets more complicated is the data layer. Every unit in the field generates gait data across terrain types, load conditions, and user profiles. The strategic logic is sound: that data trains better models, better models improve the product, which attracts more users, which generates more data. The problem is that "sound strategic logic" and "actual moat" are not the same thing.

Consumer wearable companies have tried this argument before - Peloton, Whoop, Oura - and found that sensor data was harder to defend than it appeared, because algorithmic improvements were replicable and the data wasn't as proprietary as it looked. The honest question isn't whether Hypershell should be building the data layer. It's whether data is the bottleneck, or whether architecture and model design matter more - and whether 20,000 units generates enough labeled, structured training data to make a meaningful difference.

(20,000 units at time of initial analysis; according to internal data, the number has since grown substantially.)

The SAR program is the more credible piece. 50+ search-and-rescue organizations testing devices in actual emergency deployments - sustained high-load climbing, irregular terrain, fatigue states - produces training data that no consumer dataset replicates and no lab can manufacture. Whether extreme-condition data generalizes to the average hiker is a real question, but the instinct to design field programs with data architecture in mind is sophisticated. Most hardware founders aren't thinking this way yet.

How the category got unlocked

The exoskeleton has been a near-future promise since the 1960s. Understanding why it failed for so long matters, because it explains exactly what Hypershell had to get right.

The hardware problem was fundamental: every additional kilogram of device weight increases human metabolic load by roughly 2%. A 10kg system costs you 20–30% before you've done anything useful. Old exoskeletons ran on fixed gait patterns that couldn't adapt to stairs or individual variation. "Locally effort-saving, but overall exhausting," as Sun put it at GeekPark IF 2026. A device that adds net burden isn't a mobility product. It's a constraint with a marketing budget.

Two things changed at roughly the same time. In 2019, MIT open-sourced quasi-direct-drive (QDD) motor research - power and torque densities exceeding human muscle by 10x at costs that don't require a defense contract. The second shift was AI: learned motion models replacing rigid gait programs. Sun's specific contribution was seeing something others missed. "I found an architecture that uses a single power system to simultaneously assist both legs," he said at GeekPark IF 2026. "At the time this was considered somewhat counterintuitive. But validated through simulation, it became our first-generation product - the world's first single-motor exoskeleton." That insight is what made consumer pricing possible. The M-One Ultra is its direct descendant.

He started the company in 2021, before most of the industry saw these doors opening. That's why nearly a hundred investors turned him down. It's also why the ones who eventually backed him made a good bet.

The decision that revealed the founder

In 2023, Hypershell raised $1.23M on Kickstarter - the first consumer-grade exoskeleton crowdfund, 40% of backers from the US. Then the grey-test scores came back. Nothing above 60 out of 100. Uncomfortable fit. Awkward assistance. Incompatible with backpacks.

Sun killed the product. Scrapped it, rebuilt with a new dual-motor architecture that doubled unit cost, took 18 months instead of 6, meant every second-generation unit shipped at a loss.

"If the experience isn't good enough, from the user's perspective, there is no experience at all."

This matters more than most founding decisions get credit for. The easier path was obvious and available - ship it, collect the cash, iterate later. He didn't take it. Hardware founders who optimize for short-term shipping numbers build a different kind of company than founders who kill their own product when the experience isn't right. The decisions that come later - when to cut a feature, when to delay a launch, when to spend on the right component instead of the cheaper one - tend to follow the same pattern as the first hard decision.

The second generation worked. Tens of thousands units shipped. Top-selling on Amazon in the US and Europe, and JD.com in China. Someone ran a marathon in it and swapped the battery mid-stride. Firefighters in Hong Kong climbed thirty-plus floors post-blaze while preserving enough physical reserve to perform rescue work - the first documented deployment of powered exoskeletons at scale in emergency response.

The competitive landscape

The most important thing to understand about Hypershell's competitive position is that it's not actually competing with the companies most people would name - and the reason goes back to a deliberate choice Sun made at the start.

The obvious first market for a mobility-assistance device is elderly care. Sun didn't go there. Before Hypershell, "exoskeleton" meant "medical aid" - a product you wear because your body is failing you.

A founding brand impression of medical equipment creates a permanently limited ceiling: you can grow the medical market, but you can never become something people want rather than something they need. So Hypershell went to the mountains first - hikers, backcountry skiers, photographers hauling heavy gear into terrain nobody else reaches - and built the category's first mental model around capability rather than accommodation. The downstream effect: elderly users who eventually bought the product posted about family hikes and returning to mountains they hadn't climbed in a decade, not about their mobility limitations. Same device. Different story.

That choice also defined who Hypershell is actually competing against.

Wandercraft is the most-cited comparison, and it's mostly a false one. The French company raised $75M in June 2025, has its Eve personal exoskeleton targeting the US market in 2026 - at $92,000, designed for individuals with severe mobility impairment. Its Atalante X rehabilitation system sells at $228,000 to hospitals. Wandercraft has sold 110 units total. This is a medtech company building toward FDA clearance and insurance reimbursement. It is not competing for the hiker who wants to see the view from a mountain peak. Ekso Bionics and SARCOS are in the same bucket.

The real competitive threat is closer to home: well-capitalized Chinese hardware imitators with the same Shenzhen manufacturing access and global distribution playbook. RoboCT raised $13.7M in March 2026. NAVEE showed up at CES 2026 with its Exo-Fit. Neither is dangerous today. But the pattern of Chinese hardware categories is consistent and worth taking seriously - drones, 3D printers, robot vacuums, e-bikes: once a category is validated, competition arrives fast and price compression follows. Hypershell's answer is HyperIntuition and the data corpus. Whether those hold when a well-funded imitator arrives with similar hardware in 18 months is the real durability question, and it's one the company needs to be thinking about now, not later.

The generation this company belongs to

DJI established the template that Hypershell is executing: neglected category, engineering depth over price, global-first validation, Shenzhen iteration speed. DJI didn't beat the drone market on price. It beat it because its flight stabilization was technically superior to everything else. The first generation of Chinese hardware exports won on cost. This generation wins on capability - and that's a fundamentally different, and more defensible, competitive position.

What's genuinely new about Hypershell's version of the playbook is the AI layer sitting on top of a hardware business. DJI's moat was manufacturing precision and proprietary flight software. Hypershell's potential moat is a motion data corpus that compounds with every unit shipped. The Kickstarter with 40% US backers before a dollar of retail spend is the global-validation move. The M-One Ultra from concept to production in under three years is the Shenzhen-speed move. The SAR program putting units into extreme conditions before retail scale is the brand play that doubles as real-world validation - stress-tested terrain and load profiles that controlled lab testing can't replicate. None of these are accidents, and the fact that they're all happening simultaneously at this stage of the company's development is a signal worth noting.

Sun's own read on the moment goes beyond Hypershell specifically. "We seem to be truly living in a science fiction era," he said at @GeekParkHQ IF 2026, "where things we could only imagine are accelerating into reality. Categories that people once thought impossible, or that existed but didn't have good enough experiences, can now be rethought and redesigned." What he's describing is a structural window - opened by AI and advanced manufacturing — where the gap between what's imaginable and what's buildable has collapsed faster than incumbents can respond. Hypershell got through that window early. The question is whether it can widen the gap before the window fills up.

The unit economics, roughly

The numbers that are public: tens of thousands units shipped, ASP probably around $1,000–1,200 blended across the SKU mix, gross margins 35–50%. That puts cumulative gross profit somewhere in the $7–12M range - real money, but not a business that justifies a $400M valuation on its own. Anyone who looks at those numbers and concludes this is currently a $400M company is confusing a bet on trajectory with a description of present reality.

Base case: 200,000 units at $1,500 ASP, 45% margins within 4–5 years → ~$135M annual gross profit. At 10–12x, that's a $1.3–1.6B business. Not a straight line from $400M entry, but within range if growth compounds and the software layer adds incremental margin.

The scenario that makes $400M look conservative: Ant and Meituan convert CVC positions into real distribution - exoskeleton rental in Alipay ticketing at scale, Meituan delivery worker augmentation contracts. That generates recurring revenue with entirely different margin characteristics. The standard CVC pattern is optionality without commitment, though. Whether this is one or the other is not answerable from the outside - and it's the most important question on the diligence list.

What comes next

Consumer exoskeletons are a rounding error in the overall market today - and the only segment growing at 3x the rate of the others. Hypershell has the entry point and the right brand architecture to expand from it. The outdoor positioning was always a first chapter. The Hong Kong rescue deployment, the HyperLIFT SAR program, the elderly users quietly reappearing in the user base - these are the category expanding outward from aspiration, the only direction a consumer hardware brand can expand from and maintain its pricing power.

The unresolved questions are serious. Can the data and software layer get built before the hardware gets commoditized by imitators? That window is probably two to three years, not five. Can a company whose DNA is Shenzhen manufacturing execution build the organizational capability for platform software - a skill set that has killed more than a few hardware companies that thought they could make the transition? Will the Ant and Meituan relationships have actual teeth?

Here's the honest read: the downside scenario. Hypershell remains a well-run hardware company that dominates consumer exoskeletons without ever building the platform layer - is still a good business. Probably a $1–2B business at scale, which at $400M entry is not an most neutral outcome. The upside scenario, where the data flywheel turns real and the CVC relationships convert, is something larger and harder to price.

What's not in question is whether there's a real company here. There's a real product, in a real category, built by someone who's already shown - on the record, at some cost to himself - that he'll make the right call when it's expensive. In hardware, that's rarer than it sounds. And it's the thing that all the financial modeling in the world can't manufacture.

"Different Organizations Allow Different Things to Emerge": Yang Zhilin's First In-Depth Interview

Thu, 30 Jul 2026 15:44:10 GMT

Editor's note: This conversation, recorded in November 2023 shortly after the launch of Kimi Chat, was the first time Moonshot.AI founder and CEO Yang Zhilin shared his thinking publicly at length. With Kimi K3 drawing global attention to Moonshot, we are publishing English editions of his two long dialogues with @GeekParkHQ founder Jack Zhang, starting with this one.

At the time, Moonshot had one product, a chatbot built on a 100B-plus parameter model, and three clear labels: long context, proprietary and closed-source, consumer-first. Read it as a time capsule. The transcript has been edited and condensed for clarity.

Highlights

The necessary path to AGI is a new kind of organization, not any single technique. OpenAI's success is, at its core, organizational innovation. "Only if you get the organization right can you actually walk the AGI road."
Large model innovation cannot be planned in advance. Mobile-era requirements were deterministic; AGI innovation is post-hoc. You have to try before you know.
Different organizations allow different things to emerge. Google's environment could produce a scientific result, the Transformer, but not ChatGPT. OpenAI invented nothing new, yet an industrial masterpiece emerged from combining three factors: the Transformer, 10^25 FLOPs of compute, and twenty years of internet data.
The Transformer is a new computer. Parameter count is its CPU, context length is its memory. Forty years ago people thought 500K of memory was plenty. The same fallacy is being repeated about context.
The ultimate case for long context is a lifelong AI companion. Trust and complex emotion only show their power over decades, and "an AI that has to reset its context window every day cannot do that."
The super app entry point will most likely be closed-source, because owning the model's evolution creates product differentiation from day one. This is the bet Moonshot has since reversed with the open-weight K series.
An AI-native product is defined by two datasets. Training data determines what the model can do, test data determines whether it is actually usable. Define the datasets and the product is defined.
The classical product manager pointed at one spot on the map and planted a tree. In the AGI era you mark out the whole plot and let the model sweep the field. Not design first, then build; you complete the design through making.
The metric that matters is a Moore's Law of use cases: the number of usable scenarios must double every N months, exponentially, not one scenario and one dataset at a time.
The greatest companies of the next decade will fuse two cultures: Silicon Valley's technical idealism as the drive, and the Chinese emphasis on usefulness and business models as the fuel.

Why "the dark side of the moon"

Jack Zhang: People say your company is a bit mysterious, starting with the name. What is the story behind "Moonshot AI," or in Chinese, "the dark side of the moon"?

Yang Zhilin: It comes from the Pink Floyd album, The Dark Side of the Moon. The founders all love rock music. We used to play in bands. And 2023 happened to be the album's 50th anniversary.When you look at the moon, you only ever see the lit side. You never see the back, and that gives you a strong urge to explore it. Large models feel similar. You want to explore something mysterious and unknown. It is hard, and it connects to the spirit of rock music: keep innovating, keep challenging the existing shape of things, keep imagining what comes next.We settled on the pairing of "dark side of the moon" in Chinese and "Moonshot" in English. It reflects our commitment to AGI. It might also define what kind of people we are.

Jack Zhang: I'm a Pink Floyd fan too. The album has an astronomical name, but it is really about the human subconscious.

Yang Zhilin: Right. The lead track is Brain Damage, about a person experiencing hallucination. Fifty years later, here we are trying to fix hallucination in large models.

Jack Zhang: What did you play in the band?

Yang Zhilin: Drums. The drummer keeps time and gives the whole band a frame to play inside.

Building a new kind of organization is the necessary path to AGI

Jack Zhang: How did you decide to go all in, and specifically to build a company around it?

Yang Zhilin: My understanding shifted enormously over the past few years. At first I thought language models were a tool that could improve results across scenarios. In the second stage, I thought they might be useful for many tasks. Eventually the view became: language modeling may be the only problem AI needs to solve. Everything can be addressed by making the model better, by making next-token prediction better.

In 2018 and 2019 at Google, we started training language models on thousands of chips with Transformers. You would observe so many phenomena along the way, and they kept adding evidence that this path was correct. Just keep walking it, keep finding more efficient ways to scale, and you get remarkable results on problems that used to be very hard: memory, reasoning, common sense, even complex multi-step problems. That experience left a deep mark on me. It was the runway toward founding a company.From 2020 I worked with many institutions to train large models. I was involved in some of the earliest large-scale efforts in China, including Pangu and Wudao. Through that process, I saw the challenges up close. Some were technical. Others were organizational. We found that if you use a traditional organizational structure, training frontier models is very hard to pull off. OpenAI's success is, at its core, also a success of radical organizational innovation.

So you could say I had been searching for one opportunity the whole time: the chance to build a new organization from zero. I believe this is the necessary path to AGI. It matters even more than the technical details we touch every day, because the organization is the deeper layer. Only if you get the organization right can you actually walk the AGI road.

By 2023, both the capital market and the talent market had changed dramatically. The timing was finally right.

Innovation in the large model era cannot be planned in advance

Jack Zhang: What convinced you that the organization is the core problem?

Yang Zhilin: Practice, mostly. Before this year I tried many modes: working inside a large company, working with independent research institutes, other supposedly efficient setups. None of them could produce fundamental organizational innovation.

Here is a simple example. You cannot innovate on large models through planning. In the mobile internet era, I could plan the features I wanted to build. Once a requirement was defined, it could be deterministically produced. You rarely heard of an app that a team suddenly did not know how to build. It was a deterministic event: humans encode the logic, computers execute it.AGI is different. I cannot plan that today we will fulfill a certain requirement to a certain degree, because it cannot be hard-coded or expressed as rules. AGI-style innovation is not front-loaded planning. It is post-hoc. You have to try before you know. You need an underlying machine that does many things in a systematic way.That is a fundamental difference.

Your organization has to match how you do things, so when the underlying logic changes, you need a new organizational form. The internet era produced excellent organizations, superb at things like recommendation-driven products. The new era will likely produce organizations that are superb at AGI. I think that is very likely to happen.

Jack Zhang: Is OpenAI a good template in your eyes? What did they get right, and what might not be optimal?

Yang Zhilin: On results alone, OpenAI made an enormous breakthrough. Without that company, the trajectory of humanity might be different.Going deeper: a good organization needs very high talent density, a shared vision, and the ability to focus efficiently on one goal. They did all of that extremely well.But the most essential point, and the one hardest to see from outside, is this: once you have those preconditions, how do you find a systematic way of doing things? That is the precondition for all the technology, and it is what we most want to iterate on.

Jack Zhang: By systematic, do you mean it can be replicated and scaled?

Yang Zhilin: Yes, but replication in the sense of reusing it across different problems, not copying it to another company. It forms inside one company and is very hard to transplant. But that company can use the same system again and again. Today I can use it to crack long context. Tomorrow, autonomous AI capabilities. The day after, multimodality. It is a reusable system that accumulates into your core asset. Every AGI company should spend serious time polishing this.

Google's organization let Transformer emerge. OpenAI's let ChatGPT emerge.

Jack Zhang: When people discuss OpenAI and organizational design, the debate is often bottom-up versus top-down. How do you frame it?

Yang Zhilin: The top-down frame still applies, especially for large models. Top-down is about leadership vision: can you judge what is right and worth doing, and what you should not do right now? AGI is like a lunar program, a giant system that takes a long time and many tightly coupled people. That kind of top-down design is necessary.Under that frame, what matters is whether the system lets many small units each produce efficiently, and then the top-down frame integrates the output.

Jack Zhang: So in a sense, you are running a big frame within which different directions can produce emergent innovation, because nobody can precisely define the one path to AGI today. The organization has to support emergence.

Yang Zhilin: Exactly. Look at how the Transformer came to be. Google gave those researchers an environment for emergence. Before the Transformer, the pieces already existed: attention, residual networks, LayerNorm, SGD and the training toolchain, learning rate schedules. Everything was prepared. Google provided the environment where people could freely combine them, and the emergence happened.But different environments allow different things to emerge. Google's environment could produce a scientific result. It could not produce a great work of systems engineering like ChatGPT, an epoch-defining product that pushed execution to the extreme while precisely capturing demand. Google's organization and ChatGPT simply do not match.OpenAI invented nothing new.

What emerged from it was an industrial masterpiece. It combined three things on a different dimension from Google: the Transformer architecture, compute centers capable of 10^25 floating point operations, and twenty years of data accumulated by the entire internet, which may be the internet's greatest value. OpenAI saw those three factors and provided the environment for them to combine. Out came a milestone on the way to AGI.That is my point. Different organizations allow different things to emerge. Whatever you want to emerge, that is the direction you should tune the organization toward.

The technical path to AGI is set. The product path is not.

Jack Zhang: A popular line in 2023 goes: the cards are on the table, now bet big. Meaning the technical path to AGI is settled and the game is now about who pours in more resources. Do you agree?

Yang Zhilin: The first principle of AGI is now clear: keep improving lossless compression and you get higher degrees of intelligence, eventually beyond human. There is abundant evidence. Like Newton's laws for classical mechanics, the big direction is basically settled. So that popular line has some validity.What remains is the second layer: under the big principle, how exactly do you do specific things?

For example, how do you build genuinely lossless compression over a long context? That is not simple. Even OpenAI has only taken the first step. Every subsequent step still carries uncertainty.Then, beyond technology, there is the product layer. We are still far from the superintelligent AI of science fiction, and today's products are not necessarily heading in the right direction.Every era has its greatest people and its next-greatest people. The greatest discover the correct first principles. Then a cohort of people, slightly less great but still great, solve the technical, product, and commercial challenges. That second layer is a vast open sea. How to play in it is still full of unknowns.

Transformer is the new computer. Context length is its memory.

Jack Zhang: Long context is Moonshot's specialty. With so many possible directions, why this one?

Yang Zhilin: Everyone should ask themselves what they want AI to do for them, what the human-AI relationship should be. In the ultimate form, one big capability is missing today: a much longer input window.The gap between a long window and a short one is more fundamental than I once thought. One ultimate form of AI is building long-term emotional value with a person, a lifelong companion over nearly unlimited time. Time is a critical dimension. Only over long stretches do trust, complex emotion, and decades-spanning interaction show their power. That is when AI can offer deep value to the human spirit. An AI that has to reset its context window every day cannot do that.

Jack Zhang: Tomorrow it forgets what you did today.

Yang Zhilin: Right. So think of the Transformer as a new computer with two critical dimensions. Parameter count determines computational complexity, like the CPU in the old computer. Context length is the new computer's memory. It determines how much can participate in the computation.Given sufficiently complex computation, the bigger the memory, the bigger the unlocked application space. Look at computing history. Forty or fifty years ago, everyone thought 500K of memory was plenty. Today that is obviously absurd. The same thing will happen with this new computer system. Long context, as the memory of the new computer, is absolutely essential.

Closed source, in service of a super app

Jack Zhang: This wave of model startups includes plenty of open-source players. Moonshot is closed-source with no plans to open up. What is the thinking?

Yang Zhilin: We strongly support open source. Open and closed will be complementary in this field. Open source lets developers try all kinds of applications, with stronger compliance control over data, training, and deployment, and more flexible scenarios.Closed source has its own value. The future's super app entry points, whether in productivity or consumer entertainment, will likely be built around closed models. The two approaches complement rather than conflict. The choice depends on each company's strategy. Ours is to build a super app. That is where all our time goes.

Jack Zhang: If someone wants to build a super app on an open model, buy the engine and modify it, why insist on building the engine end to end?

Yang Zhilin: Building applications on open source is a real opportunity, and the two do not conflict. But if the endgame is a super entry point, it will most likely be closed, because closed source lets you differentiate the product from day one of building the model. When you control the model's long-term evolution, you have full room to create a decisive product advantage.Applications on open models may not become the super entry point, but they can create incremental value: productization, proprietary data, fine-tuned capabilities others do not have. Both paths will exist in the ecosystem.

Also, right now we are in a technology-driven phase. A better foundation model converts into product advantage, so your base capabilities need to stay ahead of the commodity level. In ten or twenty years, when the technology commoditizes, you will need to convert first-mover advantage into more durable moats, like stronger network effects.

Begin with the end in mind: consumer is the only mode that matches AGI

Jack Zhang: So no API business, no helping enterprises deploy models. You are going consumer. The last AI wave produced almost nothing on the consumer side. Everyone ended up doing B2B, which at least reliably generates revenue. Why so firmly consumer this time?

Yang Zhilin: We are not refusing B2B entirely, but the focus and the push are consumer.For a long time, AI technology had no consumer success stories. With the new technical variable, AI can achieve results that were previously impossible, and those results can appear as new applications and new entry points, with exponential revenue growth and fast-growing users.

Midjourney, Character AI, and ChatGPT have all largely proven that AI-native apps have a real shot.And if you are doing AGI, you must choose a business mode that matches it, one that demands extreme innovation efficiency. Only consumer lets you close the loop fast. Only there can the organization form a culture of rapid iteration: updating models, adjusting the organization, and meeting user needs on a daily cadence, everything revolving around data at high speed. Only the consumer side generates that kind of energy, the kind that matches a company whose goal is AGI.This is thinking from the end backward. Co-creating with consumer users is itself doing AGI. It may even be a necessary precondition. AGI cannot be built behind closed doors.

The core is data. Without co-creating with users, you cannot get enough high-quality data, you cannot know what problems the model produces in real use, and you cannot dig deeper into scenarios together with users.

Jack Zhang: So it comes back to first principles about the goal. Without a consumer super app, enough users, and enough data, you cannot actually reach AGI. A company that does not want to build a super app is arguably not a true AGI believer.

What super app means in the AGI era

Jack Zhang: How do you define a super app here? WeChat is one because users do everything on it. Taobao is one by scale and value.

Yang Zhilin: The definition itself is not new. What changes is the value delivered. Only by providing value that could not be provided before does a new entry point appear. In the end you still need many users, high frequency, and large value created in use.But AGI has one property that makes a super app possible: the G, generality. It does not solve one class of problems. The set of problems it can solve keeps growing, so the product boundary keeps expanding. AI penetrates deeper into every part of life, the application's value keeps strengthening, and it fits the definition of a super app more and more. Generality and super app status are compatible.

Jack Zhang: That is a key point. In the mobile era, an app first reached scale, then expanded its service boundary to become general. In the new paradigm, you start with a general productivity engine, and it naturally becomes a super app. Different eras, different genes. Last era was a land grab for scale. This era, the technology engine gives you super app genes from birth.

Yang Zhilin: Well put. One addition: however general the technology, you still start from a subset of scenarios and generalize outward. And the generalization speed can be exponential rather than linear.

AI-native development: define two datasets and you have defined the product

Jack Zhang: Under the old paradigm, product managers, frontend, and backend collaborated toward a defined target, shipped in cycles, ran A/B tests. What does development look like for an AGI-era super app?

Yang Zhilin: Product development changes with the underlying technology. Mobile-era development meant clear requirements mapping to deterministic operations and fully deterministic events, built on the old computer and deterministic code. Deterministic logic gave rise to deterministic graphical interaction. Deterministic GUI plus deterministic systems: that was twenty years of internet product development.

Today the paradigm has shifted. The frontend becomes conversational UI, and more products will adopt it. The backend has been unified, to an extreme degree, into one language model. And it handles more than language. It processes all the world's information. It is encoding and losslessly compressing everything.With both ends settled, most application-layer development no longer touches backend architecture or frontend framework. There may be blends of GUI and conversational UI, but the overall architecture is basically fixed.

Most of what we call development today happens in the middle layer: data. Same interaction, same model class, different data, different product. ChatGPT, GitHub Copilot, Midjourney: essentially the same thing, differing mainly in how the data is defined.This is a massive paradigm shift. What product managers increasingly need to think about is how to build a product out of two datasets. Define the datasets and the product is defined. The training data determines what capabilities the model provides. The test data determines how usable it actually is.Before, there were no AI-native products, only AI features, so this way of working was rare. Many people with strong product sense do not know how to apply it.

Say you want to build something like Character AI, or improve on it. How do you define your two datasets? You need strong data production and processing techniques, ways to acquire data, judgments about which data is effective. We are still making these workflows concrete through exploration. The new development method of AGI probably requires a new organizational form to pull off.

The product manager of the new era

Jack Zhang: Early mobile internet ran on product managers with imagination, people who defined future scenarios by instinct. I once joked with Allen Zhang, the creator of WeChat, and he called himself a classical product manager. I like to call that quality a touch of the divine: they could not fully explain their convictions, but the convictions turned out right. Later, product management became scientific, data-driven, A/B tested. For a super app, do you want more of the divine or more of the scientific?

Yang Zhilin: You look for the balance, but long term, I believe the system will win overwhelmingly and become the mainstream development paradigm. That does not make the divine instinct unimportant. It just needs a strong system underneath. The system should be the main force.

Here is my usual metaphor. Allen Zhang pointed at one spot on a giant map and said, plant a tree here. He turned out to be a god, because the spot was right and the tree grew into a forest. A sharpshooter who hits whatever he aims at.AGI does not work that way. The industrial designer Sori Yanagi liked to say you do not manufacture according to a design, you complete the design through making. Build the thing, and the design is done, rather than designing first and building after.The old way was like scouting forever for the one spot, planting one tree, and celebrating when it survived.

Now, you glance around, say this patch of land looks decent, and the sharpshooter marks out the whole plot. AGI is your main force. It rolls across the entire field and finds every opportunity, every place a tree could grow.With a strong system, you realize that having talent-and-luck-driven judgment pick individual planting spots, out of tens of millions of possible scenarios, is extremely inefficient. The search for product-market fit in the AGI era should be done the AGI way: use its generality, use the user ecosystem, use the system, and push across the whole front at once. That is the biggest difference from the classical product manager.

Jack Zhang: So instinct concentrates at the front, on problem definition and goal selection. Getting there faster and better is the system's job.

Watch the delta, not the balance

Jack Zhang: More concretely, the product people you have hired, what do they share? Reverse-engineer the traits for us.

Yang Zhilin: An open mind and the ability to learn. Together they point to one thing: can this person iterate fast? That is the trait we value most. Not just product managers. Every role, every person, AI or not. Change is too fast now. You basically cannot predict what AI will offer by the end of next year, let alone three, five, ten years out. So everyone needs to learn fast with an open mind, then go execute concretely.For product specifically, you need strong consumer sense. Shipping version one is the easy part, because you have not yet gone through making it continuously better, or the process of defining ever more precisely what product you want. "Better" is deeply abstract. How do you make ChatGPT better? What counts as better? In which direction, and by how much? All hard to define.

A common trap for product managers is defining a pile of features, the old way. That is probably wrong now, because your features are defined through data. That is the AI-native way. And this is not static theory. You learn, then you try. What I said today might be wrong. Fine. Try it, gain something, take another gradient step, deepen your understanding. Out of that process, this era's Allen Zhang will appear.

Jack Zhang: Or not a new Allen Zhang. He was the classic of his era. The next era will produce a completely new figure, whose touch of the divine will look different. That is the most exciting part of progress. And one thing is certain about talent now: watch the delta, not the balance. Maybe ignore someone's history, but look at the increment between their past and their present, their present and their tomorrow. In an uncertain era, a large enough delta means a lot.

Yang Zhilin: Yes. Often, carrying no historical baggage is an advantage.

Jack Zhang: People in the comments are asking whether you are hiring.

Yang Zhilin: We are. Full-time and interns. We are quite open about backgrounds. AGI is a comprehensive undertaking. The technology alone is full-stack: NLP, computer vision, RL, alignment, infrastructure, kernel engineering. Beyond technology, product, operations, commercialization. Ideally, very diverse backgrounds, but one shared vision. We welcome anyone with passion for AGI, for the super app, and for the global market.

The metric that matters: a Moore's Law of use cases

Jack Zhang: Sam Altman has written about Moore's Law for intelligence, and Moore's Law for everything: a Moore's Law relationship between the cost and capability of intelligence. Do you buy it?

Yang Zhilin: Yes, and I think Moore's Law itself is going through a paradigm shift. The original version: transistor count doubles every N months. Then model parameters and compute followed it: FLOPs double every N months.For us, the Moore's Law of intelligence ultimately means this: every N months, the number of usable use cases doubles. It is an extension of scaling laws. Standard scaling laws describe pre-training: add compute, data, and parameters, and watch how training loss changes.But the metric that matters most in the end is the Moore's Law of use cases. How many scenarios reach usability? It has to rise exponentially, not linearly, doubling every N months.

You cannot use the traditional AI method of adding one scenario and one dataset at a time to make it work in that scenario. You will never get exponential growth that way.Measure how many scenarios get unlocked. With that, the search for product-market fit accelerates enormously, and you can try many things at once. Do not plant one tree. Mark out the land, and with a Moore's Law of use cases, you test the entire plot in one pass. A superfast tree-planting machine that knows whether a tree will live or die without planting it. When the scenarios multiply past a certain point, you become a super entry point.

Jack Zhang: The Moore's Law of intelligence and the Moore's Law of use cases should form a double helix. Costs keep falling, capabilities keep unlocking, and scenarios multiply.

The greatest companies of the next decade will fuse two cultures

Jack Zhang: You worked at Meta and Google. Compare Silicon Valley's engineering culture with China's. What is each good at?

Yang Zhilin: Silicon Valley's engineer culture is distinctive. Take Noam Shazeer at Character AI. They built a product with a real degree of product-market fit, and in the early days they barely had dedicated product managers. Engineers with their own ideas, fusing technology with those ideas, taking one step forward on their own. That is worth borrowing, especially in the AI-native paradigm where technology and demand have to move toward each other. Think of constructing those two datasets. Without engineers willing to take that step forward, much of it simply does not get done.

At the bottom layer, we want to absorb the best of both East and West. OpenAI runs on strong technical idealism. "I want to build AGI," business model unclear, heavily funded from the start. The recent effective accelerationism wave is another expression of it. Google and Microsoft were also born of a degree of technical idealism.Chinese culture emphasizes usefulness, thinking within the premise of a business model.

In the next ten years of the AGI era, the greatest companies will probably combine the two. The pragmatic side finds a genuinely good business model, the fuel that keeps you burning. The idealistic side drives you beyond money and usefulness, toward the simple desire to see what the far side of the moon actually looks like.

Jack Zhang: We have always been good at goal-driven usefulness. But making something useful and universal may now require some moonshot spirit: aiming at something high, hit or miss, and moving toward the deep of the universe. Exciting goals are what gather truly excellent people. Thank you, Zhilin.

The Year Elon Musk Told Me About His Space Dream

Thu, 16 Jul 2026 16:49:35 GMT

He never needed to be the one who got there. He got there anyway.

By Jack Zhang (Zhang Peng), Founder & CEO of GeekPark

*Editor's note: On June 12, 2026, SpaceX began trading on Nasdaq under the ticker SPCX, the largest IPO in stock market history. Priced at $135 a share, the offering raised $75 billion at a valuation of about $1.75 trillion. Then the stock popped, closing the first day up roughly 19 percent and carrying SpaceX's market value past $2 trillion.

With xAI now merged into the company, X along with it, and Starlink driving most of its revenue, SpaceX is finally public. Its market value, like Musk's rockets, went up and down hard before it cleared the atmosphere. The company started in 2002.

Twenty-four years later, Musk and his obsession with space are still questioned, the way they have always been questioned, and the company keeps moving toward the same goal it started with: making humanity a multiplanet species. The piece below was written six years ago, the night Falcon 9 put two astronauts into orbit aboard Crew Dragon. It is one tech journalist's first-person attempt to explain the side of Musk that most people never see. *

"Five hundred light-years." @elonmusk and I said it at almost exactly the same moment, and then we both laughed. I have no idea what was going through his head. I was laughing because I had just realized that almost no one else in that room cared about the number at all.

The number was about Kepler-186f, at the time the most Earth-like exoplanet anyone had found, about 500 light-years away. Zhang Yaqin was sitting next to me at dinner, talking with Musk about when humans might actually reach Mars, and he mentioned in passing that NASA had just found a twin Earth "around 600 light-years out." Musk corrected him before I could, almost without thinking: "Five hundred."

The NASA announcement was only days old. Musk had already filed it away as common knowledge. That was one of the moments his eyes lit up the brightest on his first public day in China.

He had earned the right to be tired. He flew into Beijing that morning on a private jet, went straight to Tesla's Beijing office, came to GeekPark's Singularity conference, sat through a brutal two-and-a-half-hour interview with CCTV, took business meetings until half past six, and then showed up at our welcome dinner. When I finally saw him he was still full of energy. You have to respect that. The man runs two of the coolest companies on the planet and clearly has a different kind of fuel in him.

Elon Musk at GeekPark's Singularity 2014 (The Innovators Summit), Beijing

Musk is not a schmoozer and not a man who enjoys working a room. He has said it himself: he is more like a slightly crazy engineer. Do not expect him to read the crowd and flatter people the way polished business types do. Ask him something that bores him, or something so off the wall he can't answer it, and he goes flat. The interface closes. Ask him the right question and you can watch the excitement move into his eyes and his whole body.

He also knows he has to repeat the same things to different people, over and over, and that sometimes he has to crouch down to meet a different audience where they are. So if you are meeting him for the first time, and you ask him something he has already answered a hundred times, something you could have found with a search engine, he will patiently say it again. This is not because he is kind, and not because he is protecting some commercial interest. It comes from an extremely strong sense of self: he thinks most people simply have not seen the right direction yet, that they need to be led, and he does not mind pointing the way for someone who is lost.

Stay lost on purpose, though, and he will hand you a cold face without a second thought. Sometimes the only thing you get is the back of him walking away.

What follows came out of the gaps: the warm-up before he went on stage at Singularity, the waiting, the walk out, and the dinner where I sat beside him. Some of the questions came through other guests I was helping translate. Because I was mostly grabbing moments wherever I could, the questions are not a tidy system, and I am working from memory, so some of the wording may be off. But I think you will still feel how this man thinks.

I asked him why he kept going with Tesla when he himself had said it would probably fail.

Someone has to push the industry to think inside a new framework, he said. He had expected, even believed, that the legacy car industry would grow an electric future on its own. It couldn't. So he decided Tesla had to build the new framework and show everyone there was another way. "It succeeding is my good luck," he said. "What I actually want is for the whole industry to change." He pointed out that Tesla was already licensing its technology to other carmakers. He was not trying to replace them. He wanted them on the right road.

I asked why he was so committed to solar superchargers.

Cost is the most important factor, he said, but in a market like China, cutting the energy lost converting coal into electricity matters a great deal too. Then he told me about a joke going around California that he clearly enjoyed: if the world ends like it does in the movies, you can still drive your Tesla, because nobody will be pumping gasoline anymore and it will run dry, but a supercharger running on sunlight keeps going for a long time. He said it as a joke. I thought the logic fit him better than he let on. I said he hadn't mentioned what I considered one of the Model S's biggest advantages over a normal car: that it can update itself remotely and keep getting better. We love that feature.

He doesn't like talking about individual features, he said, because to him Tesla is a systemic innovation, not a single selling point. But yes, you can update the car over the air whenever you want, and it becomes a better, more useful machine, more suited to you. Every detail about driving and handling that we can already see, and everything we are about to see, can be optimized and solved faster once that mechanism exists.

I asked whether he worried about the security of that cloud-plus-car architecture.

What I described was only theoretically possible, he said. No one has ever pulled it off. Not one example. They had anticipated it and run plenty of tests and found no exploit. "Honestly, I don't understand why you think anyone would do this."

He was about to go cold. From what I could tell, he treated worries about battery safety the same way: taking a low-probability event and using it to slow the big direction struck him as bad-faith nitpicking, not constructive. It has always been his style. His aggressive approach to Tesla's autonomy has been argued about in the car world for years. The Dragon capsule that just launched is another example. At the design stage he wanted rocket-controlled recovery, the Falcon 9 way. NASA flatly refused, too risky, and he compromised back to an ocean splashdown. Whether he ever truly agreed, I have no idea.

I asked why he was CEO of two companies.

"I didn't want to be a CEO at all," he said. What he really wanted was to be an engineer who could design and build products his own way. He had tried twice to bring in a CEO to run a company for him. It didn't work. He realized that if he wasn't personally in it, things went badly wrong. And obviously, if you want to do the one thing only you truly believe in, it is hard to find a stranger to make it real.

I asked why he was so set on something as risky as SpaceX.

He loved science fiction as a kid, he said, and had always thought exploring space was an extraordinarily interesting and meaningful thing to do. "I'm not building rockets because I want to go to space myself. Getting myself up there is easy." (The confidence.) But if ordinary people can't reach space, humanity stays locked on Earth, unable to explore the universe, unable to become a civilization based on more than one planet. We have to believe this is the right direction, he said, because if we can't do it, our civilization is fragile.

We have to drive the cost of reaching space down hard, he went on, and the real path is not waiting for some sci-fi technology to fall from the sky. It is taking mature rocket technology and making it reusable. He put the cost reduction at 100x, at least. "Don't you think that's worth doing?"

I said both of his companies had nearly failed and that he had put his entire fortune on the line for each, and that a lot of people probably can't understand his thinking.

"That's why I said these things are hard to get other people to do for me. I can only do them myself. I've said I might not succeed. But someone has to step up and start."

I asked if he had ever thought about going into politics. If he became president, wouldn't he be able to push his goals on the environment and space exploration even faster?

He hadn't thought about it, he said, and he didn't think he would. What he can push is a good direction through products and technology, not laws that bend the world into the shape he wants. Take the Tesla, he said. He wants you to choose it not only because going electric is a responsible thing to do for the future, but because it is genuinely a good product, which is the real reason you would choose it. He did not think being president would solve that problem. He is better suited to being an engineer and a designer.

I asked how on earth he had convinced NASA and the US government to let him get into the rocket business.

He won the trust step by step, he said. He had spent his own money on research the agencies wouldn't touch. They saw how much he had spent, how serious and committed he was, that there were results, and only then did they give him a chance to try more. He thought this was a perfectly normal process. The main thing in persuading people is not lobbying, he said. You have to actually believe in the thing, and carry it to a tipping point where everyone can see the hope.

I asked the obvious follow-up: why had SpaceX managed to bring launch costs down when NASA, with all its experts and resources, had not?

"I think the real reason NASA couldn't," he said, "is precisely that they have too many resources."

I asked what he thought of the companies offering suborbital tourist flights.

He shrugged, shook his head, and declined to answer. The closest word I have for it is a scoff. He had no interest in serving recreational demand. It does nothing for the goal of carrying humanity toward a multiplanet civilization.

Now that you have read all that, it is worth coming back to what the successful launch of a crewed commercial spacecraft actually means.

To me this was not a simple commercial rocket going up. It means commercial spaceflight has matured to the point where it can carry people, a crown-jewel mission. The space age began in the 1950s, but its starting gun was fired by competition between governments. After the US won the race to the Moon, science-driven spaceflight kept producing steadily, and yet it stayed a government program. The technology kept advancing, but to this day landing on the Moon remains humanity's highest achievement in space.

The significance of commercial spaceflight is that space exploration stops being a line item in a government budget and becomes an industry, one that can pull in more minds and more capital and build a positive loop of value creation. That story has already played out, again and again, in aviation, in telecommunications, in computing.

Only this way do you escape the paradox Musk named: too many resources, so no progress. Commerce, by its nature, has to chase efficiency and progress, and that has to come through innovation, standardization, and scale.

Commercial spaceflight is what will truly open a space age that belongs to all of humanity. You could even call it the opening of a new chapter, the moment we go from being able to touch space to actually being willing to walk toward a multiplanet civilization. Commerce is the force behind that, stronger even than curiosity, the hormone that finally lets humanity take a big step forward. Six years ago I asked Musk whether I would live to see ordinary people able to afford a trip to space. He was completely sure. "We definitely will."

It will also be a hard road. Just before this successful launch, the Starship rocket, still in early testing, blew up in an experiment. That vehicle is meant to carry 100 people to space at a time. If the Falcon 9 and Crew Dragon pairing is a small ferry skiff to space, Musk is clearly already drawing the real ferry in his head. He wants a million people living on Mars by 2050, before he turns 80. The idea sounds insane to most people. He clearly has a roadmap and a timetable, and he is running on it.

I don't think Musk will be the only person who matters in opening the real space age. He is the rider out front, breaking the wind, taking the hardest resistance head on so the rest of the pack can ride faster. More capital and more capable people will join this industry. And a new generation of space engineers in China has the same chance to become a force that carries humanity forward.

I hope the dream so many of us share, of going to space once in this lifetime, comes within reach because of this industry, and comes soon.

Good luck, Musk. Good luck to everyone in this business. And good luck to us.

How We Made the Earliest Bets in Unitree Robotics

Wed, 03 Jun 2026 22:24:31 GMT

In 2025, sixteen humanoid robots danced on China's Spring Festival Gala in front of more than a billion viewers. They were built by a company called Unitree Robotics. By the end of that year, Unitree had shipped more humanoid robots than anyone else on the planet - by unit count, at least. In March 2026, Unitree filed to go public at a valuation of 6.2 billion U.S. dollars. Eight years earlier, almost nobody had heard of Unitree, or of its 27-year-old founder Wang Xingxing. Same year, I issued a term sheet anyway. At that time, Chinese venture capital was busy chasing shared bikes, shared power banks, shared umbrellas. Robotics wasn't even a category. What follows is how that bet got made.

The WeChat Post I found Unitree through a short post on an obscure WeChat public account almost nobody had read. A few paragraphs, a couple of photos, a thirty-second clip of a quadruped called Laikago. Wang had named it after the Soviet space dog sent into orbit in 1957, the one that did not come back. It was the kind of detail that means nothing to a marketer and a great deal to anyone listening for it. I had a frame of reference. In 2015 I'd taken a delegation of Chinese founders to MIT, where one lab was running a hydraulic quadruped with a taxidermy cheetah head bolted to the front. The head was a little too large for the body, the body a little too tethered to its own cables to ever really run. The machine made the kind of noise that makes you take half a step back without thinking about it. Wang Xiaochuan — who would later found Sogou and then Baichuan AI — picked the cheetah head up for a photo. The mood in the room was that the technology might matter in a decade or two. MIT Cheetah Lab, 2015

That was the global state of the art in 2015. Two years later, a team in Hangzhou I'd never heard of was showing a smaller, quieter electric quadruped in a thirty-second clip. What struck me was that, even in thirty seconds, it didn't look like a research project. It looked like a product. The proportions, the cable routing, the way it carried its own weight - all of it had the unmistakable shape of something headed toward a customer, not toward a conference paper. I booked a flight to Hangzhou within two or three days.

The Hallway Unitree's 2017 office was on the ground floor of a nondescript building in Binjiang, the kind of light-industrial district where the neighbors are sheet-metal shops and the address is more landmark than number. I couldn't find the door. When I did, I stepped into maybe twenty square meters with no desks, no meeting table, and no chairs. Parts stacked along the walls. Half-assembled frames on the floor. Two completed robot dogs in the corner. Wang Xingxing, then 27, looked like the graduate student he'd been not long before. He ran a brief demo: the dog walked, he shoved it with his foot, it caught itself. Then we stood there for a moment, because there was genuinely nowhere to sit. Xingxing with his XDog, during postgraduate period We ended up on a couch in the hallway outside the office - not even Unitree's couch. It had the orphaned look that hallway furniture in old commercial buildings gets, the cushions slumped at the edges, the kind of couch you don't sit on so much as land on. We talked there for nearly three hours. The conversation wasn't what I expected. The usual rhythm of a founder pitch is forward lean: market size, five-year plan, the date and the number for every problem. Xingxing did the inverse. I'd float a scenario - a home robot, a rural security guard, a wildlife deterrent - and he'd walk it back. There were specific engineering problems that came first. Behind those, more problems. The order wasn't negotiable. When I finally asked the most basic question in venture capital - what are people going to do with this - his answer was: build it first, and the research labs will buy it. That was the entire answer. There was no deck to speak of, just a few slides on the technical stack designed to bring an investor like me up to speed. No vision section. No total addressable market. No plan to dominate a category. He'd thought about the use cases considerably less than I had, and about the engineering considerably more than I ever would. I'd almost never met a founder who described his own timeline more conservatively than the investor across from him. The asymmetry was the thing. Xingxing knew the ground he was standing on with a precision that didn't match his résumé - no MIT, no decorated advisor, no co-founder from a famous lab. His undergraduate thesis had been on the design of a brushless DC motor controller; for his master's, he had built XDog - a quadruped - for under 20,000 RMB, designing the actuators, the control electronics, and the gait algorithms himself, around low-cost outer-rotor BLDC motors he selected and characterized. That was the part the rest of the field tended to outsource. Almost every "robotics" startup I had seen up to that point in China bought finished motors from Maxon or Faulhaber, wrote some control code on top, and called it a robot. Xingxing had built down to the actuator. He held the machine to a higher standard than it strictly required, because the distance between good enough and actually good was the part of the work that interested him. What was missing from his pitch was supplied by the object on the floor.

There was also the matter of the name - Laikago. Laikago, 2017 I'd known the story of the Soviet space dog since I was a kid. Xingxing didn't strike me as someone who would name a product for marketing reasons - everything else about him was the standard engineering archetype, technical, undemonstrative, almost monosyllabic when the conversation wasn't about a circuit. But hidden under that surface was this small romantic gesture: a robot named after the first creature to ever orbit the Earth. I'd learn much later that as a child he'd been so bad at English that one of his teachers had told his mother he might be a little slow. A kid who couldn't pass an English test had built a robot from the motor up and named it after a Soviet space dog.

The Thesis I flew back to Beijing with a thesis. Inside Chinese venture capital in 2017, Boston Dynamics was the public benchmark - the company everyone invoked when they pictured what advanced robotics looked like. Inside the field, the real frontier was somewhere else. Sangbae Kim's Biomimetic Robotics Lab at MIT had been publishing the architectural answer for years: high-torque-density electric motors with single-stage planetary reductions, backdriveable, no force sensors, no series compliance, no hydraulics. The paradigm had a name - proprioceptive actuation -formalized in IEEE Transactions on Robotics in 2017, with MIT Cheetah 3 following at IROS 2018 and Mini Cheetah in 2019. This was the road I believed legged robotics needed to be on. None of this is to diminish Boston Dynamics. Their hydraulic Atlas did things - dynamic parkour, backflips, full-body manipulation under load - that nothing electric in 2017 could match, and a meaningful part of the field's understanding of dynamic locomotion was built on what they figured out the expensive way. Boston Dynamics Shows Off Its Humanoid Robot That Can Run and Jump But the question I was asking as an investor wasn't who had built the most impressive demo. It was which architecture would scale to a commercial product at a price ordinary buyers could actually pay. On that question, hydraulics were a dead end, however magnificent. Electric drive was the direction the entire physical world was moving: it is far easier to make a fast, light, force-controlled electric actuator than a hydraulic one. Once you have one, the rest of the problem - balance, gait, contact - becomes a matter of control bandwidth and software, both of which were getting cheaper every year. Xingxing's XDog, finished in 2015, was built around low-cost outer-rotor BLDC motors with custom drivers - architecturally in the same family as what MIT was formalizing in parallel. He didn't invent the QDD paradigm, and he wouldn't claim to. What he did was something arguably harder in the Chinese context: he took a research architecture being developed at one of the most well-resourced robotics labs in the world and rebuilt a working version of it for under twenty thousand RMB. He has also said he'd concluded hydraulic actuation could not be commercialized as early as 2013, before any of the public pivots had been signaled. His view of where the field was going was identical to mine. The difference was that he had been building toward it for four years; I had been reading about it. The quadruped form factor wasn't a dead end but a foundation. The most important things a four-legged robot teaches you - how to design a torque-dense actuator module, how to manufacture it cheaply at small volumes, how to build the supplier relationships that keep the bill of materials moving - are the same things a humanoid will need, only harder. The dog wasn't a stepping stone in software alone. It was a stepping stone in hardware industrialization, and the hardware substrate was the part with the longest compounding curve. My working assumption was that the market for legged robots wouldn't open for three to five years. The question, then, wasn't whether but who: which founder could spend that time in the dark without losing the thread. That kind of road belongs to someone more interested in the dark stretch than the eventual payoff, whose head is full of the problems and not the pitch. The hallway had answered the question for me. What I'd later admit I hadn't yet understood was how important it was going to be that Xingxing refused to chase consumer markets at all. I'd spent most of the hallway nudging him toward exactly that - the home robot, the patrol guard, the wildlife scenario. He had walked each one back. I'd treated the consumer question as a discussion. He treated it as a no. The discipline turned out to be the founder's, not mine. We sent the term sheet in November 2017. At that moment, institutional capital in China was concentrated almost entirely in software, consumers, and shared-economy plays. A term sheet for a quadruped robot built by a 27-year-old with no deck, in a garage with no chairs, wasn't just contrarian; it was outside the category that contrarian bets were normally sorted into. I could afford to see it that way for a specific reason. I'd spent the previous decade meeting and tracking a generation of Chinese founders before the market had named them. By 2017, when GeekPark raised its first fund, every LP in it was a founder I had spent years with in their own early days. Also - which I'd articulate more clearly today than I could then - this specific company could probably only have been built in China. The architecture Xingxing was working with required small shops that could machine custom motor housings at low volume and low cost, electronics fabs that would run a hundred custom driver boards without making you feel like you were inconveniencing them, and a labor market where a competent robotics master's graduate cost a fifth of what the equivalent person cost in the Bay Area. It also required Chinese university and corporate research labs whose procurement cycles fit Unitree's price points almost perfectly. The bet wasn't just on a founder and a technology. It was, implicitly, on a place. I would not have written the same check for the same founder building the same robot in Palo Alto.

What I Couldn't Have Planned For Two things, in the year after we sent the term sheet: an afternoon I'd staged that didn't land the way I wanted, and a financing structure that fell apart. A few weeks after the term sheet, in early December 2017, GeekPark held its annual founders' tea on the side of the Wuzhen Internet Conference. Lei Jun of Xiaomi and Wang Xing of Meituan were there. Wang Xingxing didn't have credentials for the conference, which is why we deliberately picked a venue outside the main grounds. I wasn't expecting anyone to write a check. What I wanted was simpler: for Xingxing to be a name they'd remember, for these founders to have seen the machine and the person with their own eyes. Around the table that day, from left to right, were: Chen Hua of Changba, Cindy Mi of VIPKID, Me, Lei Jun of Xiaomi, Zhou Yuan of Zhihu, Wang Xing of Meituan and XIngxing of Unitree. He showed up after the main conversation had wound down. He set the robot dog on the floor. The room turned toward it. It took a few steps, snagged a leg on the door threshold, and died - not dramatically, just the soft mechanical click of something giving up. He knelt down and restarted it. It walked a few more steps, but the air had already gone out of the moment. People were warm about it, genuinely encouraging in the way senior founders are to a younger one, and then everyone went home. A fun fact: Lei Jun had been Xingxing's idol since his teenage years, but Xingxing was too shy to even ask to add him on WeChat. But I still think it was worth getting him in the room, even with the robot face-planting on the threshold. You plant seeds. You don't usually get to watch them come up. A few years later, Lei Jun's fund came in at Unitree's Series A, and Meituan eventually became the largest external shareholder. Today, at Unitree's target IPO valuation, those stakes are worth roughly US$258 million and US$556 million accordingly. The second was the round itself. GeekPark's early-stage fund couldn't anchor it alone. Unitree needed between 15 and 20 million RMB, and the right structure required a larger lead investor - ideally one with hardware operating experience or a supply chain footprint. I spent the better part of the winter and spring trying to find that investor. A serious lead emerged, talks went deep enough to produce a second term sheet, then the deal collapsed for reasons that had nothing to do with Xingxing or Unitree. The shape of the problem was almost textbook: a founder I believed in, a thesis that hadn't changed, a signed term sheet from GeekPark, and no anchor behind it. The disciplined path was obvious - find a new lead and wait, or let the term sheet lapse and revisit when the structure could be repaired. Either move would protect us from carrying concentration risk a fund our size wasn't built to hold. Then Xingxing surfaced that Unitree's bank account was nearly empty and product shipments were at risk. He simply said he hoped we could move sooner rather than later. What he didn't say - what I learned from someone else - was that he'd already stopped paying himself and had been making payroll out of his personal savings. That detail didn't change the math. It made the math worse. But the math was no longer what was being decided. The question I kept circling was whether our actual judgment about the company had changed. The thesis was intact. Xingxing was the same founder he'd been on the hallway couch - more so, in fact, now that I could see how he behaved under pressure. The only thing that had moved was a deal structure. And a founder who covers his own payroll without informing his investors is not managing his investors - he's absorbing pain and asking for help plainly when he needs it. Years earlier, Zhang Tao - co-founder of Dianping, and one of GeekPark's first limited partners - had pushed me into starting the fund with a question I'd never quite resolved: You spotted Zhang Yiming early. You brought Elon Musk to China in 2014. What did you actually do for any of them beyond writing it up?

Elon Musk, at GeekPark's summit, 2014 Every dollar in our first fund came from founders who'd built their companies inside the GeekPark community before circling back to capitalize ours. Founders backing the next founder. We weren't a typical institutional VC. We weren't supposed to be one. So the question wasn't whether the deal made sense given the broken structure. The question was who we were. The discussion went around the same loop several times. At some point I stopped, sat with it a moment longer, recognized the thing I couldn't sit with, hit the table, and we wired the money. It wasn't, by any honest accounting, a clean piece of investment reasoning. In the second a decision like that gets made, it doesn't feel like an investment. It feels like the answer to a question someone asked you, years before, about who you were. The Loop That Worked My confidence at the time had included an assumption that GeekPark would help raise the next round through our founder network. We never got to find out. Xingxing outpaced us by shipping product. By late 2018, Laikago was in the hands of paying customers - exactly the customers he'd named on the hallway couch: research laboratories and universities that needed capable hardware at a price a research grant could cover, machines open enough to be taken apart and built on. Scientists and engineers buying tools. The revenue was small but it was a loop: product, cash, better product, more cash. Xingxing was running the company without marketing spend, without influencers, without the apparatus of growth - just the object on the bench and the iteration of it. The non-obvious part of this strategy was that academic customers are not just buyers. They are the most leveraged distribution channel in robotics. A PhD student who buys your quadruped will spend three years finding edge cases in your firmware you never would have caught, will publish papers benchmarking your hardware that function as free marketing to every other lab in the world, and will graduate having become fluent in your SDK. Boston Dynamics' Spot, at 75,000 dollars and closed-source, could not compete for this audience. Unitree's machines, at a fraction of the price and open enough to hack on, captured it almost by default. By 2024 Unitree had over sixty percent of the global quadruped market and a meaningful fraction of the next generation of working robotics engineers had been trained on its hardware. That was the loop compounding. It's what allowed the company to spend years building motors, supply chains, and gait software during the period when no one in the consumer or enterprise market was prepared to buy a legged robot anyway. It was a moat he was widening, every quarter, while the field was empty.

The Catch-Up In April 2024, Boston Dynamics released a farewell video for its hydraulic Atlas - a montage with music, the long goodbye a company makes for an architecture it has outgrown. The next morning, it unveiled an all-electric replacement. Boston Dynamics Puts Its Hydraulic Atlas Robot Into Retirement Eleven years of hydraulic engineering had been relegated to the museum. The architectural argument Xingxing and I had agreed on in the hallway in 2017 had been validated by the one company in the world that had spent the most resources arguing the other side. By that point, Unitree was the company global robotics labs benchmarked against. Xingxing had moved into humanoids with unusual speed, and the reason was the one I had not fully understood in 2017: he had spent six years industrializing actuators. The motors, the drivers, the reducers, the manufacturing relationships, the iterative cost-down across thousands of quadruped units - that was the moat. Unitree is now heading for an IPO at a valuation in the range of 42 billion RMB - roughly 25 times trailing revenue, on a profit number that contracted by half in Q1 2026 as price competition intensified. At this multiple, Unitree is trading on a humanoid market that has not yet arrived. The honest reading is that the early thesis electric drive, quadruped as foundation, founder who could survive the dark stretch has been validated. Whether the humanoid market arrives on the timeline the price implies is itself a bet, and not one I would make with high confidence today.

But the bet isn't the only reason this story matters to me. In 2014, GeekPark brought Elon Musk to China and put him on stage next to an almost-unknown 31-year-old founder named Zhang Yiming. Neither was a household name yet. Three years later, we did it again: a 27-year-old in a Hangzhou garage, working on a category that didn't yet exist, in a year when every other check in Chinese venture was going to shared umbrellas. What was different the second time was what we could do about it. In 2014, we had a stage. By 2018, we had a community, a fund, and the willingness to wire money to a founder burning through his own savings to keep his company alive. GeekPark has never been a typical investment institution. The fund is one expression of something older underneath it: a community that tries to recognize the right person on the right trajectory before the consensus does, and tries to be useful in whatever way the moment requires. Sometimes that's a stage. Sometimes it's an introduction. Sometimes it's a wire transfer on a day when the math says wait. Unitree as a company today is something none of the people in either of those rooms could have drawn a straight line to in 2017. The thing that began as a conversation on a hallway couch, and limped across a doorway threshold at an awkward afternoon tea, became larger than the bet that started it. The dog stumbled. We wired the money anyway. That was the bet I made.