"I finished college at 20. I grew up in a rural area of China with limited resources, but I developed strong curiosity early about how the rest of the world looked. Right after college, I spent two and a half years traveling the world with no money. I earned some income from university research work, bought a one-way ticket to Europe and the US, and lived by couch surfing, hitchhiking, and camping across unfamiliar places. I often did not know where I would sleep the next day or what I would eat, but things always worked out. The best and worst part was the same thing: dealing with uncertainty. As a founder and entrepreneur, you face uncertainty every day, every moment." - Rocky Yu, Founder and CEO of AGI House


"I did not need to live that way. I had family support in any situation, but I chose to do it out of curiosity. I wanted to understand the rest of the world, what young people like me were doing, and what they truly cared about. I do not believe that visiting a place for a few days is enough. I deliberately spent long periods living at their level to see and experience life as it was. I carried a 70-liter backpack with a tent and sleeping bag and traveled around the world." - Rocky Yu, Founder and CEO of AGI House


"We have seen many stories, and one word comes up often: resilience. You need to be extremely resourceful and resilient. People talk a lot about talent, but the world is not short of it. What differentiates people is who goes the extra mile to make things happen. When you realize, as Steve Jobs said, that the world is built by people no smarter than you, your perception changes. You realize you can be anyone and build anything." - Rocky Yu, Founder and CEO of AGI House

Rocky Yu, Founder and CEO of AGI House, joins Jeremy Au to unpack how early curiosity in computer graphics led him from engineering and startups to building one of the world’s most influential AI communities. They explore why talent density matters more than scale, how AGI House emerged during the pandemic as a mission-first experiment, and what it takes to turn deep technical conversations into real companies. The conversation covers Rocky’s journey from academia to entrepreneurship, how dinners and hackathons sparked breakout AI startups, and why AGI should be understood as a system of applied intelligence rather than a single god-like model. Rocky also shares his views on resilience, uncertainty, and how young people and parents should think about work, purpose, and opportunity in an AI-shaped future.

02:00 Early fascination with computer graphics shaped Rocky’s path: Curiosity about how computers generate realistic images pulled him into computer science long before AI was mainstream.

06:06 The pandemic triggered a mission reset: Isolation and deep conversations about purpose and intelligence sparked the idea that later became AGI House.

08:12 Talent density became the core design choice: AGI House prioritized curating elite researchers and founders over scaling a broad, open community.

12:32 Invite-only dinners and open hackathons worked together: Private discussions built depth while hackathons surfaced raw, unproven talent who later broke out.

15:29 Resilience comes from knowing why you build: Rocky explains that founders who love status quit early, while those driven by curiosity endure hardship.

17:21 AGI is a system, not a single god model: Intelligence emerges from many specialized agents improving through real-world deployment.

29:02 Learning to live with uncertainty builds founders: Traveling the world with no money trained the mindset Rocky later relied on as an entrepreneur.

Jeremy Au (00:00) Hey Rocky, really excited to have you on the show. Here you are, technical superstar, CEO, and founder, and now leading AGI House and a community. So good to see you in Singapore. Rocky, can you tell us about yourself?

Rocky Yu (00:12) Yeah, sure. First of all, thank you for having me here. Actually, I'm here for a conference and then got a chance to talk to you. My name is Rocky, founder and CEO of AGI House. It's a community of AI founders and researchers. As the name suggests, it started as a "hacker house," but now it's much bigger than a house. We have a company operating the community. The mission is to accelerate humanity's transition to AGI. At the same time, given that most of the important AI companies in the early days, back in like 2023, started from the community, we recently set up a fund investing in early-stage and pre-seed founders.

Jeremy Au (00:46) Yeah. Fantastic. Let's go back in time. I understand you were in Beijing, you were an undergraduate, and you chose to study computer science. So why did you study computer science instead of business or something else?

Rocky Yu (00:55) There's a lot of coincidence; I also was determined or destined to do certain things. And of course, curiosity drives much of my life decisions. I think a big part of the choice to study computer science was that I was fascinated by the idea of how computer-generated images could look so realistic. That was more than 20 years ago. I was fascinated by that and my research and activities were around computer graphics at the time. Like Toy Story—how are they generating all those things using computers? I was a big fan of Steve Jobs and the studios.

Actually, it leads to what I'm doing today because right now, the thing that powers the whole AI innovation is a company called NVIDIA. At that time, NVIDIA was less known than today. They were doing one thing: selling computer graphics cards. It's the piece of hardware that handles rendering computer graphics because a lot of those algorithms are very computationally expensive. So you need dedicated hardware to do those renderings for the graphics. I was doing research on that. Then back in 2007 or 2008, they first released CUDA. That's where they started to repurpose computer graphics as a general-purpose computing platform, which now runs all the AI. All the training runs on that. So I was very early and very lucky to be in that whole wave that later paved the way to what I'm doing now.

Jeremy Au (02:19) And so there you are, learning computer engineering. In the early phase of your career, you were a technical leader and engineer. So tell me more about that.

Rocky Yu (02:27) Yeah. From the experience in computer graphics and research, I went to a school in Beijing and then I went to the US for graduate school. I met a lot of professors doing other research. I knew that my goal was no longer being a researcher because I believed that the alternative was a better way to really push the technology forward. So I dropped out of my PhD program and started a company called Digimetrics, actually using CUDA to decode digital video and audio to detect bad frames. It was like a digital SaaS product back in 2009 and 2010. And then we sold the company and I did a few other things. So that was my first contact with the digital infrastructure for AI stuff.

Jeremy Au (03:06) And you were building these companies. What was it like building those companies as an early employee and as a technical leader?

Rocky Yu (03:12) I think the most important thing is the love of the technology. New things come out and we see the potential of what this can do. One thing we realized at that time is that video and audio processing were still very computation-intensive. The hardware was very limited, and this new paradigm—the CUDA infrastructure and parallel computing platforms—could really unlock a lot of potential. That's where the product ideas led to the things we were building. Looking back, from an investor's perspective, you know that was a very big bet on this whole building paradigm. I feel like as an investor, that was probably the best time to get into NVIDIA, 10 years ago.

Jeremy Au (03:53) And I think what's interesting is that from there you actually went on to not just be like a CTO, but you became a founder and CEO of Piki. Could you tell me more about that?

Rocky Yu (04:01) Yes. In Silicon Valley, there's a phrase, "serial entrepreneur." You build a company, but you don't stop there. You want to do new things and keep exploring ideas. So I did a few other things in between, but I think the most relevant thing was back in 2015 and 2016 when I had the idea—we saw a lot of real-time social data from Twitter and Instagram. One thing we observed was that if one location has a lot of images, videos, and text about that geotagged location, probably there's something going on. So that led me to think: can we use AI to figure out what's going on and "write" news? It's basically citizen journalism. ByteDance, which now owns TikTok, was starting around that time, but they were doing something much simpler—they were using algorithms to make recommendations. We thought, "Oh, that's not cool. We should use the AI to write the news itself."

But looking at the technology then, we were a little ahead of ourselves. Today, actually, it's doable. If you look at a lot of reports that are generated, they are human-readable. But the good thing was actually meeting Greg Brockman, who started OpenAI. I think everything leads to something. I still remember we were at his apartment watching the movie Ex Machina and trying to see what the future of AI looked like. Right now, look at today—all those people in that room have become very important people in the industry. At that time, we were very nerdy researchers getting together to talk about the future of AI. But now they have changed the world fundamentally. We got to really tap into that community of people.

This leads me to what I do now. Basically, over the pandemic, because I'd been running a company and pivoting ideas for a few rounds, I was asking my friends—many of whom had worked in AI for a long time—how far away we were from AGI. Over the pandemic was the best time to think about those questions because everyone was locked at home asking about the meaning of life. So we did a lot of soul-searching. The idea for AGI House came from that: first, we need to get together, invite best friends to live together, and host events inviting the best people in the industry, whom we happened to know early on.

We were very lucky because one of our friends, Andrej Karpathy, was leaving a house that was then called NeoGenesis. It was more of a party house at that time. Andrej was in between positions—he was leaving his job at Tesla where he worked with Elon Musk on FSD (Self-Driving) to rejoin OpenAI. He didn't want to run the place because it was an expensive house. So I got a chance to take over. I rebuilt the community from the ground up with the mission of bringing together all the best researchers in AI. In late 2022, the people who really knew how to train foundation models were very limited—maybe a hundred people really understood it. A lot of them hung out at AGI House. And then ChatGPT came out and people realized, "Oh wow, AI is no longer the AI we were talking about before." VCs were trying to understand what was going on. So as an entrepreneur, we saw a large opportunity: we brought in the capital, we brought in the big VC firms like Sequoia and Greylock to sponsor events, so they could bring capital and talent to the community.

We hosted the AGI Dinner Series where about 35 founders and researchers would have a four-hour long discussion. On weekends, we hosted the AGI Hackathon Series. We would invite 150 people to our place to spend a day building early prototypes. You see a lot of companies come out of that—Perplexity, Cognition, coding agents, LangChain, LlamaIndex, Pika Labs, Coframe. That's how we built our initial reputation.

Jeremy Au (08:56) Amazing. And I'm just so curious because you made a transition from being a founder to being a community leader. I'm curious about that in-between moment. When you were between jobs, what were you doing? Did you meditate?

Rocky Yu (09:12) I think two things we did were very consequential. Number one, as I mentioned, this happened during the pandemic. We had a very small core group of best friends—five or ten people hanging out regularly because of social isolation. Right after we had the vaccine, we started traveling every other week, searching for what to do and how to see the world. At the time we started, the mission was more important than the company. I wish we had set up our VC fund at that time; we probably would have had the best-performing fund because every other event, three or four unicorn founders would come out. It was a crazy density of talent. I spent effort to recruit the best talent for each event. I would go to professors or AI labs at Stanford and MIT, inviting the best people from OpenAI and Google. At that time, we didn't think about the business part of things. We were mission-driven. After our GPT-4 hackathon sponsored by Greylock, all the VCs and founders were calling us. That's when you realize what it means to have product-market fit.

Jeremy Au (10:51) No, I love what you just said—you were searching and then you started doing it for fun, and it became huge. I like that phrase you had: "talent density." You spent a lot of time making sure the right people come.

Rocky Yu (11:04) Yeah. A lot of people are building communities. For me, after building AGI House, there are lessons I've learned. The core principles of building a community start with your mission. We set a very ambitious goal, and that attracts the best people because the smartest people want to solve the toughest challenges. But they also want to hang out with people they like. So it's very important to curate a diverse group of ambitious people with technical visions so they can bounce ideas around. My happiest moment is hosting those smartest people every week. I work probably 20 hours a day, but the joy is incomparable.

Jeremy Au (12:11) How do you prioritize time? For a community, you want the best people, but there's also a lot of young, rising stars and raw talent. How do you make that curation choice?

Rocky Yu (12:24) That's why AGI House has different categories of events. For example, the Private AGI Dinner Series is invite-only where we actively search for established people. But then there's the AGI Hackathon, which is application-based. We leave room for a serendipitous network so that young, raw talent has a chance to be seen. We ask people to recommend the best raw talent. We see a lot of Stanford PhDs and people who haven't proven themselves yet, but they get a chance and end up doing amazing things.

Jeremy Au (12:58) You mentioned everyone was on a sofa watching Ex Machina. Now they are all very successful. How do you nurture people to be successful from your perspective?

Rocky Yu (13:09) We like to use the phrase "unleash human potential." For a certain group of people, you have already made your way to a certain place—you are hungry and driven. Given the right resources and environment, we believe there's a chance you can do amazing things. We have a residency program. When we started, we invited ten residents who actually left their jobs. Six of them were Stanford PhD students. They had the drive. We help spark the entrepreneurship journey through hackathons and dinner discussions. Pika Labs and Coframe ended up raising millions within a few months. The idea for Coframe came from an AGI dinner conversation about the future of UI/UX for Generative AI. He built the initial prototype at our hackathon and hired his first few engineers through our events.

Jeremy Au (14:29) Amazing. And what do you think prevents raw talent—who are hungry and ambitious—from succeeding? What causes them to fail?

Rocky Yu (14:44) We see a lot of stories. It comes down to one word: "resilience." You need to be extremely resourceful and resilient. The world is never lacking in talent. The differentiation is people who really go the extra mile to make things happen. Steve Jobs said that when you realize the world is built by people no smarter than you, you change your perception and realize you can be anyone.

Jeremy Au (15:21) I love that quote. How do people build resilience? Do you coach people on that?

Rocky Yu (15:29) I think motivation is the most important thing. Why do you want to do this thing? Some people think building a company is "cool" or they want the title of "CEO." But then they run into trouble and just give up because that's not the right mindset. Being a founder means you have a vision to build a solution and capture a business opportunity. You have to keep asking yourself why you are doing what you're doing. That is where you build up resilience.

Jeremy Au (16:09) I'm changing gears a little bit here, but you're using the term AGI—Artificial General Intelligence. How would you describe AGI from your perspective?

Rocky Yu (16:21) That's one of the most heated discussions. With the hype of the last few years, people were debating if it was imminent. For us, our understanding is that AGI is not about one giant model doing everything. All models are essentially statistics and neural networks. True AGI is a gradual process where intelligence is applied to solve real-world problems. We have "agents" now—coding agents, marketing agents—creating value. But it's still very early. Eventually, we will get to AGI, but it's a gradual adoption by society. It's not just training one big model.

Jeremy Au (17:29) You made me think of the movie The Matrix. When we use the word AGI, we often think of one "God" that does everything. But in The Matrix, you have Agent Smith, the Keymaker—different roles within the same system. How far away is AGI?

Rocky Yu (18:04) Real-world adoption is still in the early phases. If you look at the model itself, Ilya Sutskever, former Chief Scientist of OpenAI, mentioned that LLM pre-training as we know it is hitting a ceiling. A lot of the progress is now in post-training and other directional things. It's still unclear what the best path is to true AGI. From an application layer, there is still humongous economic value to be created. It's hard to give a definite answer—it could be five years, or it could be 20.

Jeremy Au (19:18) One thing about AGI is that folks in Silicon Valley are debating it, but when you go to the Midwest, or Asia, or the Middle East, people are asking, "What is AGI and how do I use this technology?"

Rocky Yu (19:40) I was having lunch with a Permanent Secretary and a Chief AI Officer here in Singapore. When I talked about AGI, they shook their heads. Outside of Silicon Valley, there is a reality check. Most early innovation happened in Silicon Valley in a very small area. But now that we are moving toward the "agent application" layer, we see companies coming from different regions. This is one of the best times to talk about the diffusion of technology because foundation models are starting to stabilize. People are really starting to focus on real-world adoption.

Jeremy Au (21:18) Are you looking for founders or partners in this region?

Rocky Yu (21:21) We want to tell the story so that people get hands-on experience. We have a company called AGI House Labs. Our mission is to accelerate the diffusion of technology by working with enterprise customers. We have a community of the best AI talents, and the rest of the world has the real-world problems. We believe that is where the magic will happen over the next few years. And of course, from the VC side, we are always looking for founders to do amazing things. Even though we are mostly focused on Silicon Valley right now, founders can move to the Valley to build and then grow their startups anywhere.

Jeremy Au (22:30) There's a lot of debate about AI and jobs. Steve Bannon said in an interview with The Economist that he is a "Luddite" because he's worried about job losses and wants a regulatory slowdown. How do you think about AI and jobs?

Rocky Yu (22:55) It is inevitable. If you look at previous industrial revolutions, they always repurpose people from one job to another. In the AI revolution, traditional businesses that used to require tens of thousands of people will realize they are no longer needed. This is already happening. The way to look at it is: how do we redistribute wealth generated by these corporations? We might see a shift to a three-day or two-day work week. Humanity started with seven days a week, then moved to five. I don't agree with slowing down because that slowdown is driven by fear. You should never make decisions based on fear. This technology can solve hunger and poverty. Imagine robots running 24/7 to solve hunger and basic materialistic deficiencies. We will learn to deal with the issues rather than fearing job losses.

Jeremy Au (25:05) Any advice for Gen Z or parents of Generation Alpha who are concerned about AI?

Rocky Yu (25:28) When we started AGI House, we used as much AI as possible. Gen Z is a very lucky generation—they should expose themselves to it as much as possible. Every generation brings new opportunities. They will create jobs for themselves. For parents, every technology is a tool that can be used for good or bad. How we manage the degree of adoption will be a process of learning. Some people might indulge in it for a while and then figure out how to treat it as a tool that is part of their life in the right amount. Humanity figures things out as they grow and evolve.

Jeremy Au (27:35) And my last question: could you share a personal story about a time you've been brave?

Rocky Yu (27:39) My life is driven by curiosity. I finished college in a rural area of China when I was 20. Right after college, I spent three and a half years traveling the world with no money. I bought a one-way ticket to Europe and America and used Couchsurfing and hitchhiking. I didn't know where I was going to sleep tomorrow or what I was going to eat. I learned to be resourceful. I chose that because I wanted to understand what young people around the world really cared about. I traveled with a 70-liter backpack and a tent.

Jeremy Au (28:57) What was your favorite memory and your worst memory?

Rocky Yu (29:00) The favorite and the worst are the same thing: dealing with uncertainty. As an entrepreneur and founder, you are dealing with uncertainty every single day. Some people just can't deal with it; they want certainty. When you learn to deal with that, that's where you distinguish who you are and build character.

Jeremy Au (29:30) Awesome. On that note, thank you so much for sharing.

Rocky Yu (30:21) Thank you for having me.