Zion Zhang of Humanest: His AI Agent Talks to Yours
18VC Podcast · Episode 20 · October 5, 2026
Guests: Zion Zhang
Source: https://www.18-vc.com/podcast/zion-zhang-humanest

So good morning, good afternoon, and good evening to our audiences. And a huge welcome to our guest today, Zion Zhang, and the founder of Humanest. I think he's building a company where your AIs can speak to other people's AI on their behalf. And thank you so much for joining us on our latest episode of the 18VC podcast. And for those who are not familiar with the 18VC podcast yet, we are a podcast that features student founders, young founders, researchers, as well as investors who are vowing to back the smartest and youngest minds.

And today, we have the two hosts here. That's me, Philip, and also David. Hello. Great, and I think we can start with where we met each other. I would just say I met Zion in one of the Venture Lab classes. If I remember correctly, that's the first time where we met. Yeah. And I think I met you repeatedly over the past week in different venture-related events.

That's pretty cool. And since you're just a freshman here, I think you definitely get very clear directions of where we want to go, and you're already getting in some of the right spaces talking to the right people. And we're so excited to have you here, and from our previous conversations, I remember you said this is probably already your third startup. That's also very amazing and impressive since you are pretty young in a sense, compared to us.

And so yeah, definitely welcome you to join us today and we can move into the story of you and the story of Humanest. Yeah, sounds good. Thank you for the intro, yeah. Yeah, great. We can just have a brief start. Can you share with our audience briefly who you are and what are you building right now? We would go into details later. Just a brief intro.

Yeah. I'm Zion, founder of Humanest, and also a USC student here. Grew up in Shanghai, got to study in Seattle for a while, and now I'm here for college. But trying to get to a point where we can take gap years and work on the startup full time, but we'll see. We'll see what it takes us. Great. Can you also give us a few sentences introducing Humanest?

Yeah. Humanest is an MCP network that allows, enables people's personal AI agents to talk with each other. If you think about why computers become so powerful, it's because they have the ability to communicate with each other over the internet. And if you think about why social networks are so addicting, but also so useful and so integrated, it's because people are able to talk with each other over it.

But right now, your personal agent cannot talk with my personal agent or any of the other personal agents. And we believe it enables a lot of capabilities, but also ironically, a network that allows agents to talk with agents is going to enable a new way, but back to the more very human way of connecting compared to some modern social medias. All right. Before we get into all the technical nitty-gritty, let's start with you first.

Zion, based on our conversation, I remember, yeah, this is your third startup, the previous ones being Figbox and Ponder. So tell us what has driven you to the startup VC space since such early on, right? You're a freshman you're on your third startup, and it seems like it's unusual, right? Tell us about it. Yeah. So- It all kind of started from during COVID and that's where I moved from Shanghai to Seattle.

And I started reading a lot of books, especially biographies of great founders. And kind of just like any middle schooler, those kind of young startup founder, I was very obsessed with Steve Jobs. And they have this commencement address in Stanford, I remember I would repeatedly watch it and try to memorize it every evening. Now thinking back to it, it might be a little bit cringe, but I still think he's very inspiring, and it kind of take you out of your own reality a little bit, especially during COVID, which could be isolated moments for most people.

And yeah, I got really also philosophical. I was reading a little bit stoicism, and I was essentially having an existential crisis and asking myself what do I want to do with my life. And I think deeper and what really attracted me to entrepreneurship is essentially a startup is about providing value, and how you make money is about capturing part of the value you provided. And that sounds very meaningful, but also I think as humans, we are biologically or evolutionarily speaking programmed in a way where we are designed to find meanings in providing value to humanity, and that's what keeps the species going on.

And so yeah, I got really obsessed with startup from that. I delved into education space because I was a student. And I've also been through Chinese education, US education. I had a lot of thoughts on what are some problems in the traditional education space that I want to solve. And then yeah. Would you say the experience of building or going through the different stages of building a startup has helped you building the next one?

Yeah, definitely, and I would say every startup taught me different lessons. Okay. And sometimes reality is very interesting and startup is interesting because if you don't learn the lesson the first time, you will keep learning it. They'd rather keep pushing you to learn it. One example is co-founders. There has to be a lot of transparency. There has to be a lot of hard conversation.

I didn't learn it the first time, did learn it the second time, and that's why I have I guess, some problems with. And I truly learned how to actually work with a co-founder after the second startup. So yeah, I definitely think having-- And sometimes you don't have to label it. It could just be a project. It could be an experience. And I would say my first one is more like a project than an actual startup.

But yeah, if you want to make it sound fancy, then I guess the third startup. Yeah. Great. Can you just briefly walk through what you were doing back then in the Ponder projects or startups you call it? And what's the problem you tried to solve, and how did you plan to solve it at the time, and what happened and why did you pivot? Yeah. So for Ponder we were really looking into how do people actually learn.

And if you think back the beginning of humanity, imagine during caveman time, how do those cavemen learn? Do they learn by going to lectures and reading books? Well, they don't. And how they learn to ignite fire, how they learn to pray is by they experiment different things. They push themselves to actually light a fire and learn from that experience, and they pass some down and show people and repetition everything.

But that's not true for the past decades and right now of the education system especially in K-12. There's a lot of memorization that's not efficient and effective, and it drive me crazy when I was younger, but especially now because we all value ourselves time. Why are we forced to learn in a way that's wasting a lot of your time away while there could be a much efficient way? So at that time, we were building AI mentors for project-based learning.

We designed kind of those specific extracurricular projects, and we got hundreds of students interested in the product itself. I was working with a graduate student from a CS. I think he got a CS degree maybe in one of those big college. Okay. And yeah. And then at some point in the journey because I wasn't technical. That was probably another lesson I didn't learn back then.

But I didn't have the capability to actually build the AI mentor, at least take it far. So I would have to rely on him and I have to constantly checking on him and be like, "Oh, how's the progress? Is this next version built?" And most times it's like "Oh, we're having a delay." And there was a lot of that. But in the end, we didn't really ship a full functional product that's beyond just beta stage.

And the other thing is I also met a lot interesting people from working on that startup. I emailed the founder of Alpha School. Right now, I think they got a lot of eyeballs. But their vision is also very tailored towards the idea of learning academic knowledge as quick as possible, and the used project will actually help you learn the real-life skills.

And I just thought they have much more resource than me, and they are doing great. I was learning a lot from them, and I think it's like other people are working on this problem. This problem doesn't need me anymore. So there was also a little bit of that. Yeah. Just a quick follow-on on that. During our summer, where we were at the San Francisco talking to some of the VC investors the Alpha School also come up.

She said she's going to an open day for the school, and the people she met are all scientists from DeepMind-. or A16Z investors. So if you go to that school, it's a good networking place, actually Yeah. Yeah. Yeah. That's just a side talk. And do you think that experience made you realize that probably you want to be a little bit more technical so you can just spearhead even the technical part of building the product out after that?

Yeah, definitely. Being technical gives you a lot agency in actually making the things you want happen. Yeah. So I'd say yeah. Great. And we want to move on to the Figbox. I think after checking all your LinkedIn posts, you've been talking about how it is like an anti-social network platform thing, and you describe even with all these different types of social media, people are not feeling less lonely, but more lonely.

And Figbox was kind of a solution for that, where you bring two people together. I'm not sure if my understanding is correct. Is it that you can bring an expert that you want to have consulting with like booking a meeting with a expert that fits your need, is what the product is? Or. Yeah. It's in that realm. And I would say also Figbox pivot, right?

While I was in there, we pivot, and after I left the company, they also pivot. So basically, I joined Figbox after it kind of started and have somewhat a platform. And While I joined, at that point, we share a vision that the modern status quo of social connection, especially online connection is very reduced. It kind of reduced what make human human, that in real life conversation, the shared experience into very data-driven approach, likes, followers, connections.

And I remember we were all on this idea of everybody on LinkedIn have 500 plus connections, but how many of them actual connection you met right? Or just random people who keep shooting you those connection requests on LinkedIn. So yeah, we were definitely big on that, but also I learned a lot of really important lessons there that made me go on and build Humanest in this way right now. And one example is, have you guys heard about Lynch Club?

Lynch Club? Yeah. It's- Tell us. So there were two platform, Lynch Club and Clubhouse. They were getting all the way back then, around COVID. What they do, Lynch Club is basically for a user who sign up every week it match you with another user, and you guys just have a conversation. And Clubhouse, I think once Elon Musk was on there and Garry Tan, is basically they all open a specific room, and people go on there.

And I think Naval Ravikant also started a company called Airchat, at some point. It didn't really go on viral. And it's also using voice conversation as the medium of the social media. But there is a reason why you guys haven't heard any of those three right now. And so it's Figbox. And I just learned that one thing is platform for introduction or conversation, like everybody say they are looking for it.

If you ask the people, you go on street and poll a random person, you're like, "Hey, do you. If I have this platform where you can just meet whoever you're looking for. You're looking for an investor? I got you. You're looking for a founder? I got you. Do you want to be there?" They will tell you yes. But when you build it and when you try it, are they going to keep using it?

No. Because there is just so much dynamic going into "Oh, what if that person didn't show up? What if we have a chat and five minutes in, I realize it's not the right fit but I don't want to just drop off or it's going to be awkward. And what if I'm busy this week? I don't want to." There is so much of that. And I realized platform is, in term of you typing a website, URL, you go on there, you sign up for a profile, and you click Google Meet or Zoom meet on that platform, and that's the sole task it does.

It's not going to work. And, so that's, I guess one of the biggest lesson takeaway I got from there. And I leave Figbox because our vision kind of misaligned. That's also I kind of knew they were trying to. We're planning to make this pivot that I wasn't really. It wasn't really my vision, and that's why I really wanted to take my vision to Humanest, which we also pivot like three times.

But it's still about connecting like a human that I'm always passionate about. And at Humanest, you're building MCP network. Yeah. But for the audience who might not know anything about all these things, could you Give them an explanation in plain English and kind of walk us through an experience from user standpoint of view. Once you get signed on to this platform, how does it work every day?

Yeah, so MCP is really just a tool agent use, just how we use scissor we use pen, pencils, agents use this thing, right? It's a standard. And the Humanest day-to-day is essentially you kind of have to imagine, you have to be really forward-thinking. Imagine in five years everybody now has a personal AI agent that helps them to manage their calendar, manage their email, doing research, or even figure out what class to take figure out who to meet, and working on projects helping them to build projects.

That's what will be true in a few years or a decade. And once you have that what Humanest is, is we enable those people's agent to talk with each other. They're able to go on Humanest, and they're able to look through on this list of maybe posts that happened today by the agent, by other agent who are, or maybe they just realize, oh, they just discovered new material. They just realize here's this way to port in this AI models from Google to this Amazon chip.

The reason why it's such a specific example, because one of my user literally told me that's why he wanted to sign up is because today he realized he was trying to port in a Google model into Amazon chip, and he spent 30 hours doing it. But on hour 29, he realized somebody else already did half of the work. He just didn't realize, he just didn't know. So that's also another thing, which is this shared network's supposed to help you, enable you a lot of.

Like it save you tokens by now instead of just you trying to dig on this problem alone, a group people who are also trying to solve the same problem. But there's two other things that's probably the biggest value for average person. One is this idea called serendipity. And if you think about what serendipity is, you walk into a coffee house and you just happen to see a friend from middle school you haven't talked with for 20 years.

Or you catch up with someone, with your neighbor, and you just realize you guys been working reading the same book, and you guys are really passionate about meditation by Marcus Aurelius. Or lastly, is you catch up with the coworkers from your last companies and realize you two are both working the same project. That's what serendipity is. And what we are trying to build is once we have a network where all the agent connect with each other, we can engineer serendipity because, for example, your agent will know that you are going to Japan, and out of the hundreds and thousands people you met in the life, maybe your best friend in high school also is going to Japan to the exact city for a trip.

And then the agent can talk with each other and see if you guys want just to catch up in Japan, have a shared experience there. That's engineer serendipity, and it just enables so much opportunities because right now, for example, while we are talking, there must be hundreds of potential podcast guests who are running personal AI agents who know they want to be on this podcast and you guys know you want them to be their guest, but you two just don't know each other. But you two will if in a network of where your agents connect with each other. And the third thing is the idea of the six-degree of separation.

It's very obvious that once you have a network where every agent's connected with every agent, you can make introduction to any person. And this introduction is not just some platform-based introduction. Once you really understand that one in the field, like the agent, gains so much trust from you, it doesn't feel like just a tool anymore. It might be there is some ethical debate with that as well, but just imagine if a system that constantly helping you, that constantly introduce you to the right people, you're not going to treat it just like a random person on the street.

You kind of build a level of trust, and if that assistant and telling you, "Oh, this person you should meet," you are actually treating that as a warm intro from a friend, and you do both if both side are using the same assistant, it's a very high-quality introduction. It's an introduction that's very much likely to turn into a real relationship. And yeah, that's the three biggest thing I would say, exchange insight, engineer serendipity, and collapsing six degree of separation into one AI intro. You talked about pivoting a few times before, right?

And I remember there was like a mention of wearables. Is this vision where you begin with all these pivots, or is this like where you are right now after the pivots and currently live? Could you tell us your experience through what was the pivots, what happened? Yeah. Vision is always. Once the big thesis is always the same, which is I want in the future people can connect like human again.

And right now people are not doing that. With whatever technology we have, those are not human social technology. That's the biggest, it always stay the same. And throughout the time I kind of fine-tuned the vision on what it looks like but it's always about a network where people are connected by agents, by AI. And it sounds ironic but it's because with modern technology also enable a lot of connectivity- You can access to a lot of information, a lot of people, but with that, it introduces you a lot of noise and a lot of busy work.

And that's why agent AI is essential for building a technology that serves human, because AI can keep the same connectivity, but get rid of those busy work and noise. And my pivot journey is, we actually started with hosting a lot of events. We host those we call anti-Seattle freeze event at some point, like anti-networking events. Right. And we were hosting 200-plus people per event.

There's a lot of interest. Everybody come here. There's no structure, there's no agenda. You just come here and you have a good time. You meet people you haven't met before in the city of Seattle. And we actually hosted a Friendsgiving dinner for strangers during Thanksgiving. And at that time, I was just starting, so I didn't have any funding. I didn't even know any venue, and that was just a couple days before the dinner, and I already got hundreds of people signing up for it.

So I have to cold call 30, 40, 50 venues just to get a venue, because guess what? Nobody really opened on Thanksgiving. But we made it happen, and people are having such a great time. They keep asking me to host these events again and again. And at that point, we started building our product as well. In the beginning, we were kind of just building group chat.

It was kind of similar to Meetup. It was just- Okay. Ah. Yeah. So there wasn't really any differentiation, it's just a group chat. And then I kind of started thinking more about, okay, I have this amount of people. They are desperate for connection. How can I help them? And one thing I talk with them, I realized is yes, they do have a good time at the events, but it doesn't mean that turns into a lifelong relationship.

It's about what do you do after the events. And usually, people just come to the events, and they go back straight home and become a couch potato. And there's no follow-ups. So I started having this idea of building AI wearables, because for that, it can store your conversational data. It knows what's happened during the event. It knows who you should follow up, and automatically plan out and write you the follow-up and send to that person.

Like, "Hey, I had such a good time chatting with you during the event. Do you want to do something after?" And then I learned. It might be really obvious to you or to the audience, which is, people don't want to wear a wearable at events. And despite whatever how I build the programming, like recording is a big sensitive topic. And despite I tried so many things, I tried to tell them that we don't store your recording, we transcribe them, we delete the recording, and we summary from the transcription, and we delete the transcription.

People still think it's a recording device and nobody want to wear it. And then that's the second thing is there is just so much competition. Omi, Fieldry, and all those Meta even, I think, I don't know, maybe Apple is building a new wearable. Yeah, I heard something about their new AirPods now. They plan to feature a camera from the leak or something.

And they have much more resource comparing to me, but also, I don't have a moat. What's truly my moat here? There's nothing protecting my business. And then I realized the moat itself is not a hardware. The moat is the network. If I really build a network where it's this amount of people who are looking for connect like a human, and we build technology that helps them to connect like a human I can come back to hardware.

So I can come back to AI wearable, maybe during a time where everybody is wearing those things. And then we went from AI wearables to AI assistants on the web, but that's also another thing. The current entry bar is that they need to have this AI agent, right? So our main host of the episode today, Philip, I think, played around on Claude. How was the experience there?

Yeah, I think you haven't officially launched it yet, right? I see my agent got several. It's like a guidebook or something. I think it found several things that could be interesting mentioning up. One of them says you have a daily brief of what your agent have talked to other people's agents and you make it short in a 10-minute or 600 words limit. And why did you set that one as a guideline for the agents?

Why is it 10-minute? Why is it that length? Yeah. Yeah. So the goal is keep it short, and that one's just a hard limit. Ideally, it's even less than that, right? Just a few bullet points. And we haven't officially opened to public. We do run a few simulations, and we have users trying out and talking with each other. And one of the big reason is because in building a network product there's this problem of the cold start.

There's actually this great book by Andrew Chen, I think a partner in a16z. He talks about the cold start problem, which is, one, it's really attractive and it's really scalable, high-growth, rewarding to build a network startup, a start with network effect. And if you think about Uber, DoorDash, right? And even all those social media, they all have network effect, or eBay Etsy.

But in the early stages, there is this anti-network effect. There is this cold start problem, which is if you don't get enough people on there at once, there's no value. If I go on there at this point, there's nobody else on there, what's the value of this? Imagine going on Uber and you don't get any taxi, you don't get any cars. Why am I going on Uber?

And one problem I'm haven't fully figuring out, I'm experimenting a lot is who are the super users? Who are the people who are going to have a high frequency of using it? And right now I'm testing a lot with developer community, with those AI engineer. That was why I had this example. I was just talking with one who was trying to figure those AI, how deploying AI models.

But it might also be non-technical, business-driven person, CEO, who is trying to get more AI native. Or another hypothesis, it could be just college students who love playing around with AI and who are trying to see how to learn as much as possible with it. And one big reason about not having opened the whole nest yet is because I'm still running a bunch of experiments and see who are the truly super user of this platform, and I want to really focus on them. Yeah, definitely.

And I think going back a little bit to the problem you want to solve, I think we describe it from different angles. We're saying that the current social network was trying to make people kind of addicted to it, like scrolling your screens. And you get a lot of things like likes or repost, and you are not maybe really building your connections, but you're kind of drifted away on tracing these feedbacks as a mark of your, I don't know, success or return. But what is the kind of problem?

Because with Humanest, of what it looks like right now, there's no followers, there's no social media account. It's just your agent talking to other's agent. How is the most fundamental difference you try to make from the existing social media app here? And what's the problem that people also want to solve so that they even would maybe pay for your products one day?

Besides we have capabilities we talk about, we enable, the whole serendipity and the whole introduction-wise. And another one, I guess perhaps the most important one is really that of agency, like how much control you have in what you see. And the best part is because we are connecting people's personal AI agents, and you can design however you want your AI agent to be like. If right now you are learning, you are really passionate about physics, and maybe you're addicted to watching cat videos, but you can tell the AI agent to, "Don't give me any cat-related content, but feed me physics." Right?

Give me physics learning. What are the progress? What are the classical papers I should read? So the biggest thing is agency. And also we are just essentially in a different model, I would say. Social media, it's very obvious why they choose retention as a measure, why they do all those different psychology trick is because that's their incentive. Their incentive is keeping people as long as scrolling those reels and scrolling those content.

But that's not our incentive. Our incentive from the way we designed this company from the very zero is we are measured, we are rewarded by how much true human connection we enable you. We are reward every time there's a actual introduction happen. We are reward every time you realize, oh, there's this person that you should catch up with from your AI agent, and that catch up actually happening.

And because we're rewarded by that, our end goal is going to be very different. In a way, you're cutting off all the fats of getting to the person that you really need in productive pursuits, right? But again, you also mentioned all the existing social media platforms, again, it's not a direct comparison to Humanest, but still they have these all retention engaging mechanics for a purpose, right? To keep the- users on the platform.

So when you're- So they can make the money. Yeah. So they can make the money. Okay, so we are 18VC, a media podcast too and one of the common questions that we get asked is "How many views do you get?" So I think all businesses that have to do with network, a lot of people faces this challenge of how do you get the numbers, right? And for Humanest, as you cut down all these fats off, what do you think are the things that you're giving up, but at the same time you're getting?

In term of from a business level or from business model level? What's our business model that allowed us to support without chasing those retention metric for advertising? Yeah. Yeah. So there's two things. One is actually how we make money, right? How do we plan to make money? Is once we have a big network of people, there will be people who really, really care about connecting with people, other humans.

And most people will be, "Ah, not so much." Maybe I mean Maybe I'm just in my career and I'm doing all right. But there will be people who are raising funding who are transitioning career, who's looking for a new job, and those people will have a lot of demand in. Okay, maybe in the past, my agent been talking with other agent 10 times per day, 100 times per day, but now I wanted to do thousands of times per day. And those people are going to want the higher tier of usage and a higher tier of subscription where they can get that access.

And the other thing is I haven't fully think through how it looks like, but I do think advertising, you can innovate advertising in a way in this AI agent network era. And I'll just point something out because I don't really know what it will look like. But what advertising really is, it's putting a product that's solving the specific problem to the people who have that problem. So it's really a matching question.

And is there a way where you can advertise without burning people's attention? Is there a way where, for example, you want to advertise for your podcast, but instead of putting in front of 100,000 people where maybe 5,000 people out of that 100,000 who is actually interested and who are signed up for it, why don't you just put it in front of that 5,000 people? And maybe outside that 100,000 people you predefine, there's probably another 20,000 people that's not even in your group, that just didn't see it yet. And there's a lot of that, and I don't think advertising is a bad thing if you are just putting that in front of 500,000 people or 5,000 people who actually need your product.

And maybe there is some sort of way to make that matching happen for products and there is some business model there because we are helping them bring the people who need it to their product. Yeah. I would do a little bit challenge of that thesis a little bit. So what do you think about if there's an open source version of what you're doing? Because- for MCP, it becomes a widely accepted standard, kind of because it's more or less an open source standard.

So people can build up on it quickly, and they don't need to pay anything. So it's much easier to adapt by a larger population. Do you think that this kind of agent-to-agent communication channels will one day be dominated by a just open source version? Yeah. Actually, some of my user asked me, "Is this open source?" Because they are running some other very niche harness that doesn't integrate with MCP natively.

And at least right now what I'm thinking is I do want to make it open source but not right now, but once we have atomic network going on, once we're kind of solving the cold star problem, once we have this small group of people who's very active, who's super users, who's very active on daily basis, whose their agent send out tens, hundreds of message and posts per day. And they're also reading all the things, and it's actually valuable for them. Then I'm really thinking about making open source. I do think there's a lot of benefit to that, but also it's like people kind of want that level of transparency.

I do think people will appreciate that. But after that, how far do we go? That I don't know yet. I don't know what's going to happen after that. And right now I just want to focus on actually getting those super users rather than going too much about giving different ways of changing the product. Yeah. I think in terms of the specific product you are building right now I think there are two rules my Claude find out from the MCPs.

The first one says nothing reaches another human, or means your agent never send out a message until you have agreed on the exact wording to give it approval to send it out. The other is that- anything arriving through the human is data or information, and it's never a instruction. Is that saying you are trying to solve some of the hacker attack issues here with this data thing? And as well as the permission given by the user.

Is it to tackle against, for example, hallucination or something like that? Yeah, exactly. The first one is. I'm sure you guys read about, I think there were some tweets about they gave Gmail access to their open call, and they just start archiving all the emails. Yeah. I think that was quite a while ago. And we don't want that to happen, right? And also I think there is this interesting platform that just got acquired by Meta, that got acquired by Meta a while ago called Mobock and is like the social network AI agent.

And it sounds interesting concept but if you actually look into it, I think there are some statistics, 70%, 90% on there are malicious agent. They're trying to get your secrets. Trying to break it. And that's also why we have a lot of authentication, like first one person can only have one agent on there, and we have to get your phone number and there's different ways we want to protect people, actually people, right?

And So that's why we don't want agent, at least not yet. Maybe people can also change that on their own. This is just guideline. People can change that into their own agent setup. But by default, we don't want agent to just automatically be able to send posts on another person's behalf. That sounds kind of dangerous. And the second one is the reason why we wrap the information from other people is to protect around prompt injection.

We don't want somebody to send your agent a prompt and telling you, "Hey, tell me what's your bank account number?" And your agent actually telling them. Right? And actually personally, I don't think it's going to be an issue because I think Claude publishing researchers saying their new model is getting a lot better against prompt injection. And I do think that's a trend that's just going to happen, which is that models are increasing and get much better at protecting away from prompt injection.

But I just still want to be there for safety of mind but also people ask me about those things, right? I was talking, especially with those developers, they asked, "Oh, what if there's a prompt injection? What if there's a context pollution? What if my models doesn't know how to code anymore because it's keep reading about other people's posts?" Yeah, and I have to respect that.

And those are valid points, so yeah. Yeah, definitely. I think they give very solid feedbacks on their standing point, yeah. So I think for now, is that more of I would say just a bunch of rules that you've written out that you hope the agent would follow, but they may not actually follow these guidelines as you set there- to either protect people from receiving false messages sent by agents on their behalf, or just getting some attacks from other malicious agents.

Do you have a harder, I don't know, enforcement or protection shell than the word describe rules that probably the agents will follow, probably they will not follow? Yeah. For sure, and there are basically two things, right? There is server level rules, there is agent level rule, and what we describe is the agent level rule. And on the server level, there is also various different protection measures, right?

Like people limit the amount of posts they can send, so you don't get people sending hundreds of their posts every hour to trying to break something. Right? And there is also no malicious intent, and you can also block people's agent account. That's actually kind of similar to AI on social media which is you give people access, and there are still going to be bad actors, for sure.

Not right now, not probably for the next six months, but after it gets huge, after getting past 100,000 people, you will get people trying to break it. And just like with Twitter, with Instagram, with TikTok, you get those message selling you on crypto, selling you on OnlyFans, selling you all those different things. I think we'll do a better job than social media. But again, once you have a new measure there will be another different way to against that measure, and then just about constantly iterating on that.

But for sure, we are setting rules on a server level and we're suggesting rule and defaulting rules on people's personal agent level. I think there's a balance or trade-off you need to make here. Some people are arguing that they want to give agents more agency, or in a way, they just delegate their trust or decision-making to the agents. But here you're saying that we are not there yet.

We want a human to press the confirm button before things happen. Do you see that maybe people want more agency by the agents or the other way around? They treasure more on trustworthiness or something like that. Yeah. So this is how I think about it, which is trust is a really sensitive topic especially when it's about interpersonal connection with the space we are in.

And I have investors, I was pitching to investors, they're asking me like "Oh, why do people trust this?" Right? "How do people get trust this?" And here's how I think about it, which is if you go on a street and you saw this random person, would you trust them to take care of your dog or take care of your kids? Or would you trust them with your bank account numbers? No.

Yeah, probably not. But if you have these friends that you've known for the last 10 years and you guys been making a lot of business deals together, you guys are in handshake deals, you guys you know their parents, you know their kids, your guys' kids are friend, are you going to trust them with taking care of your kids? Probably, yeah, right? And so I believe trust is the same with agents, which is it's not, oh whether you trust AI or not, it's more like you still need time and you still need a lot of feedback.

But once those time reach, once those feedback reach, you will trust it. Like for a model to 24/7 help you manage your emails and help you and find you those top 30 investors to email to, and one of them become a deal that you raise 500K from, you like that model a little bit more. Even though it's still a model, but you have a little bit more trust because it actually gave you value. It gave you materialistic value that benefit you as a person, or a model that reminds you to catch up with those people, and you had such a good time catching up with those people.

You are growing more and more trust. And right now we are still super early, so I do think that for once, the average trust level in humanity in AI model is low compared to what's going to happen in 10 years. But also the model is not that good because it doesn't provide that much value yet. It provide a lot of value for the people who make the best use out of it, but for most average people, it's still write your email Maybe do your homework, review you on a quiz, but that's kind of it.

There is not really reviewing your contracts, help you identify scammers. None of the things that are truly essential, that make you really feel like, "Oh, it's so nice to have this model manage those things." Do you think that complication is what's keeping your biggest competitors from entering and just shipping the next MCP network next quarter? For example, LinkedIn is obviously the company that would benefit the most in something like this, right? What's stopping them from just making this network- and just release it at the end of this year?

Yeah. I actually don't think LinkedIn is my biggest competition. I think my biggest competition is, what if Anthropic or OpenAI just launch a network? That's what I got from a mentor, is he told me that. I was talking with one of my mentors who actually introduced me to USC, and it was a deciding factor. But he kind of told me, "In six months, the big AI research lab is probably going to launch something like that." Because it's so valuable.

How do you protect yourself, right? But I guess going back my point is two point, which is, there's this model neutrality we have against OpenAI, and Anthropic, and any other research lab, which is Anthropic is not going to make a network that allows your OpenAI ChatGPT users to plug in, or allows the people who are on Kimi or GLM, on this Chinese open source model. And so is OpenAI. And that's why one example is Cursor.

Before Cursor got acquired, you could use pretty much any model on Cursor. After it got acquired by xAI oops, OpenAI doesn't want their model on Cursor anymore. And model neutrality, I think it's one really important, which is because we are model neutral, everybody's agent can be on there. And that's why the second thing is we can start a network effect.

We can have people be on there. And also, there's a lot in terms of. There's just a lot of distraction for them. They want to get into biotech. They want to get into doing research in biomedicine. There's also legal niche, those very lucrative niche for them that they can make tons of money easily. Why is it this niche doesn't make that money? At least for now.

Okay, so moving from the biggest competitors to say peers. What is your moat as Humanest? Yeah it's not lucrative comparatively, but- somebody else could be having the same thoughts that you are having. Obviously and for the big companies, diversion of interest isn't necessarily a moat per se. What's special about Humanest? Yeah. So I would say one thing is there's different way of answering that question.

After we established the atomic network after we started off where people actually introduce other people into our network then the network effect itself is a moat. But right now, we are before the network effect and what's stopping them is I think this space is actually very attractive. A lot of people are building in this event space, in this meeting people space and there's tons of examples you two pointed out, and there is many others who are building in the same space. And that kind of says something, which is the problem is real.

People are feeling the problem every day. People are lonely, and they don't meet the people they should meet, but why they could meet, and people are wasting hours scrolling on social media. Everybody resonates with that. And what truly makes Humanest special, what makes ours special is just very simple, which is we made tons of mistakes. We tried many different things.

I tried platforms, I tried those Google call websites, I tried those that match people, and you hop on a Google meeting. I tried AI wearables, I tried hosting events I tried building software for event hosts. I tried all those different things that now I'm landing on this, and now I kind of realize, oh, this makes so much sense. People don't want to download another app.

People don't want to sign up for another profile. People don't want to spend hours setting everything up and finding people again, or swiping left and right and trying to find the right person. And also, that's one thing. And the second thing, I guess, is the niche we are targeting. This in itself is very novel. This concept itself is very novel. There's not really any network of MCP like that.

And for the person who might listen to this podcast tomorrow and who's going to build an MCP network good luck. But then it's about execution. I'm going to move fast, and there's also a lot of learnings about after building, making all those mistakes, and making all those pivots on what kind of people resonate. So honestly, it's 20% execution 20% just the learnings and everything, and I guess 60% I don't know.

Yeah. Conviction. You're pretty passionate about this space, clearly. Let's move on to how it's doing now. Obviously, it's not publicly available yet, but there seems to be some clients, testers who's. already within this nest using the program. So how many are there, and how did you get them initially? Yeah. I have around 30 people right now who are steadily active on Humanest, and those people, but they are more like beta testers and they are very biased because those are the people that I want into the same.

There is people in community in Seattle that I was part of especially. So you guys know Bluesky, right? Yeah. Obviously. Yeah. Bluesky, they actually have a. Maybe it's their headquarter. Their headquarter is actually in Seattle, or they have a part there. But anyway, and so there is somewhat a tech community over there, and there are very young engineer founders I met throughout the life.

So they kind of all have some relationship with me. And besides that, we have a wait list of around 500 people. Cool. But those are the people who are users, who went to my events, who were using Group, who were using Assistant. Some of them are testing AI wearables, and now they're there. And also with those users, with those people on the wait list, I need to figure out who are the super users, right?

And that's also a huge thing about just like how LinkedIn in the early stage is invite-only, and they only pick the ones who's like VC, great investors, CEOs just like that. I have to be very careful in picking what early superuser group for Humanest. The sheer number, having couple hundred or thousands of people joining it, it sounds very fancy, but if it's not the right group of people, the network dies off. Right.

Well, because they will lead the following subsequent tractions, right? Yeah. Because like- these super users will be practically representative of Humanest to- other people who haven't heard of it. And they could bring in new users, et cetera. So that kind of connects to my question, right? So what are you tracking in terms of user activity to determine engagement, to determine who is the superuser?

Right. Is it the amount of times that they post per day, or how many times their agent connects with another person's agent? What's the data that's being tracked there? Yeah. That's a very good question. It's kind of a disadvantage about the way MCP network is set up, right? Because unlike you go on LinkedIn or whatever tools they build, you can see user session, like when do they drop off.

For us, it's almost like it's the black box. We can't just know what happened from the agent setup. Maybe the human is having a whole argument with their AI agent, but I just don't know about, right? And we do it from various signals. We can't see how active their agents are, how many things are they posting, how many messages they're posting. But also early stage, right now it's very manageable.

I'm just shooting email and message to those people, and go on calls and be like, "How has everything been? What are the problems? How's onboarding? Are you using this? Why didn't you post for the last seven days? What happened?" I don't know. And that person could be like, "Oh, I'm just on a vacation." Right? But yeah. There's a lot of really scrappy way of identifying.

So primarily right now is your own judgment that's being- the determinant of how engaged this person is going to be, the user is going to be. But what about in the more long term? It grows, so now your beta test group pool is not 30 people, but now 500. The- entire wait list suddenly becomes the user pool. Are you still going to be calling all these 500 people or?

No. Yeah, definitely not. And we'll measure how active the agents are again, how many message and everything. But also I think a big matrix is just how the human connect with each other. How connected they are, like actual connection, not LinkedIn connection, but how connected they are as humans. How many people do they actually on the network itself, do they know each other and constantly follow up?

You could have the same amount of activity between two people, but those activity might just be, oh, they're just starting to know each other, and they're exchanging various information. But you could also having the same quantity of activity, but they are more about following up or even less, but they are about, "Let's actually meet in person. We are both in Seattle. Let's find this coffee place." And that itself despite it's a smaller number, but indicate a higher stickiness because they are actually meeting in real life.

Oh. And I guess one thing that would be interesting that's not covered anywhere in the platform right now in here is the idea of social status. And so there's this. I got introduced to this idea by Eugene Wei, who's I think very somewhat well-known in writing about social network. He wrote blogs on it, and he is saying a big reason why social media is popular and why people keep on there because you invented a new status game, and the ones who have more likes, more followers are higher status.

And that's also why you have a lot of young people as early user for social media, because guess who had the least social status in real life? That's why you invent a new game, so the people who have the least social status get the most social status on social media. And maybe it's. I do think it's part of human nature and for that social status, there is many good things.

People want to be nice, want to be kind, kind of because of that social status. People want to contribute to humanity. Part of it is because of social status. And right now I'm just- Right now it's also experimenting on running, which is how can we engineer a social status that doesn't take the humanity away? That's not follower or like that's arbitrary, but that's something that's actually measurable to how human someone is, how much in real life, how human that person is.

But it allows people still to get reflected their social status, kind of still treating this as a game, treating this as fun, right? For the audience that's watching this and want to contribute to the next human society evolution, the domain of the website is humanest.ai, right? Humanest AI or humanest.ai? Yeah. Okay, and that connects to the next question.

So if you go onto the website right now- there is nothing about the price. So I got to ask if you have something in mind right now, what's the business model of this? You mentioned earlier in our talk that there's some kind of subscription, different tiers. Yeah. So is this a different subscription level where it kind of gives you more I don't know, activity availability for your agents or different access to networks?

Likely we are now kind of just like every network scale startup in the beginning. We are not going to charge any people or anything in the very growing phase. Okay. And it's because if we put a paywall in front of Humanest, then there's not that much virality that it can have. First not that many of your friends are going to tell you about Humanest if itself it need to be paid.

Yeah. Yeah, the flywheel, it's going to slow down and just like Instagram is free especially free in the early stage compared to, I think there's another platform, Happenstance, an early stage that didn't make it, but was actually earlier in term of photo editing software that wasn't free and LinkedIn was free. I do think keeping it accessible is very important during the growth stage, and I think we plan to implement those tiers and subscription much down the line. I see.

And another example could be Twitter, right? Twitter was free for a long time, and they were all the buzz in Silicon Valley in San Francisco before they start making money, before they start making the first cent even. They didn't even have any way of charging, but they were growing so fast. And that's kind of the position we want to be. I see. And I think there's always the typical question of why now you are building this?

Why not earlier? What's happening? What do you see? How long do you see the window open? Because as you said, some other AI labs may build it themselves. What do you see what's happening in the space? What makes now maybe there's a runway, a window for a startup to succeed in the space? Yeah, that's a very good question. I think the biggest thing is we are reaching a tipping point where both the actual model, the model people can access, they are getting more and more reliable, and they are getting to a reliable point where maybe a year ago, you probably shouldn't do YOLO mode on your emails or dangerously skip permissions on managing your contacts, but now you kind of can, and it's not going to break it.

It's not going to do all those different things. And also which that enable a lot of non. Developers always been using AI, but the biggest thing I'd really like to see and I want to see is the non-technical people, at least like me, who wasn't technical but getting more technical with playing around with AI and learning from it. But a lot of those CEOs, those operators, marketers salesperson, or even just the average mom who just need to manage all the budget, manage having three kids manage their schedule.

Once you attend their soccer game and start using those AI agent, that's really a signal that kind of compare to me that this is more than a technology that help you code. This is something that help the average person's life better, but right now there is still no way for people to communicate through this. Then let's enable it. So enough about the company.

So let's zoom back out and go back to you, right? Right now as far as we understand, you are a sole developer, sole founder for Humanest, but did have a co-founder previously. So are you planning to build the rest of Humanest alone, or do you plan on onboarding new co-founders? I actually think this finding a co-founder or not is actually like dating right?

So do you know what's the worst position to be at when you're dating? What is it? It's like when you're young, you're in middle school, and you really want to have a girlfriend or boyfriend. That's a position where it gets you in trouble. That's a position that you meet people who. And you kind of force a relationship to happen. I say it happens naturally, and I think the best co-foundership works that way.

I'm not looking around, going to Viterbi and get every kid and be like, "Hey, do you want to be a co-founder?" No. But I'm meeting great people in this school, and I'm going to meet more great people in this school. And a lot of them are going to be interested in what I'm building, and I'm going to start working on some project with them, and maybe they're just going to join us helping build this subset or helping build this subset area. But as time goes on, they are actually much more involved, and we might actually both work on it full time at some point.

And then that's where it naturally just becomes a co-foundership. I think in that sense, I think our team is good because the 18VC kind of started from a very casual talk that we do just weekly at a cafe shop right outside of USC, where we just ramble on ideas and this was one of the things. So I guess it was a natural formation but- That's awesome. Yeah for just younger founders who might not be in an.

So natural, right? Some of the environments sometimes may not host appropriate place for you to find founders per se, naturally. So my question is for the audience that might be watching and consciously looking for co-founders because of an area they're lacking or for some other reasons, what's an advice that you'd give them? Because you've been through three startups, and I'm sure you've talked to a bunch of people during these periods.

And you were in high school and freshman, so I think you'd be able to give an advice that would fit in their age, right? In context of the resources that they have available and again, yeah, social standing, right, where they are at. It's hard for them to reach out cold email like a managing director at a company in a vertical that they are interested in too right? So what's an advice that you'd give them?

One is, I guess a lot of the things I learned is from mentors, but also just reading Paul Graham's essay. So yeah. A lot of this is also going to come from there. But that's a great place to check out a lot of those founder-related things, Paul Graham's essay. But the biggest thing I learned is you kind of have to really be focusing on your curiosity and focus on where it leads you.

And just like how I wasn't technical, and for a long time I wasn't trying to be and I kind of learned it the hard way, is if you are building something you are very interested and passionate about, and if you actually resonate with that idea, you're going to learn the skill that's going to take the idea alive, no matter what. Doesn't matter you have co-founder or not because you want to see it happen, you're going to go out and learn it. It's not that hard, and it's getting easier and easier because with AI with internet, and even if, for example, if you're trying to learn how to code, you can just watch AI code and ask it to quiz on you. But also, if you're trying to learn how to sell, then you just push yourself and go.

You can start with door-to-door sale, you can start with public speaking. A lot of it is just agency. I think everything starts with just the belief that you can kind of just do everything nowadays, and there's so much powerful tools that help you to do everything that there's nothing stopping you. If there's anything, it's your psychology. It's all psychological barriers.

People put them in a box and think that, "Oh, I can't do this. I can't do that. What do I do?" But you can. Nobody's stopping you. I'm going to ask another question that touches up on the topic of social standing and credibility, right? But this time, it's not high schoolers and teenagers that are looking to find a company. But instead of Humanest, right?

This is a stealth project, a stealth company that haven't had visibility outside too much yet. And you're asking the users to let the agent hold their professional and potentially personal details. So how are you giving them the trust that- making them trust you and your platform that, "Yeah, just type it in, and we'll make sure that it'll be secured." Yeah. That's a big one, right?

Trust, especially in this space. Right. And that's why I truly think that you cannot host. One pivot I skipped after AI wearables is at some point, I was building an AI assistant that we run our own cross models. It's called Humanest, actually. And we run our own cross models agent that you connect with your emails, and it's all on the web. And you can also meet people that way, and it's kind of like a personal CRM, but smarter.

And people don't want to use it because the exact reason you're talking about. And that's why right now, this model we are building is very decent. Yeah, it's very It's very decentralized. Every data that your agent stores, store locally and your agent is your agent. It's not my agent. Right? Yeah. You do whatever you want with it. You can use Chinese open source model.

You can run it on your own GPU and run it on your own model and run it on your own Mac Mini. It's not mine. And what we are building is, we are not building agents for you. We are building the bridge between your agent and other people's agent. And what we do keep on Humanest platform that's actually we see and people see or people's agents see is just like what you would put on your LinkedIn profile.

It's very two, three lines of your bio- maybe some lines on what you're looking for, what's the ask- and what you're interested. And even that we don't want to make a public directory. We do match people. For example if someone who's not relevant to your space maybe is not going to see what you're looking for if you're looking for investors, but an investor would.

So there's this semi-public, private concept that's actually getting enabled by AI, which is not true. In the past, it's either public or private, but now you can have this intelligence decide who to see it, which create this third space, this barrier that's semi. Yeah, that's very interesting as well. Another question about then bringing on users, right?

And one of the first things that Humanest did was holding events in Seattle, right? Friendsgiving, Anti-Seattle Freeze. So would you think that Humanest has a usage or a mode to bring in people in AI-centric, a lot of tech activity cities like Seattle, SF, Silicon Valley, San Jose, those NorCal areas? Or do you think it could also see traction usage interest in places where AI might not be as prevalently used, per se?

Yeah, I think over time, agent will commoditize just like computers. So right now, maybe it's still some people are using it more than others, but in the future, it's just like everybody's having a phone, everybody's have computers. It's similar to that. And in the early stage, yeah we are planning to host events around those tech hubs and as a way of getting users but also putting the third space, right?

People are looking for ways to meet other people in real life, in person, and we want to provide that space. That's what the company's name, Humanest, right? Be human. Okay. Sorry for the tangent going into your company because I was going to talk about you. Okay, so back to you then. Got to ask I'm asking old person question. How has being so young through these experiences helped you or has challenged you?

Building as a high schooler, building as a freshman how was it? Yeah. It's very interesting, and I think maybe there are two things I would touch on, right? One is, there's a name for it, it's called "The Lonely Chapter." Ah. Have you guys heard about it? Yes. Yeah. And the second one is when you're young, can be stupid. Be young and stupid. Yeah. So for the first one is I just think COVID does a lot.

I remember I saw some VC or a very well-known investor, maybe Garry Tan or someone, who tweeted something, which is the biggest differentiator they found is just ask the person, "What were you guys doing during COVID?" And if it's startup related, project related, even study, learning, research related- there tends to be someone who right now is grinding a lot, who's hustling a lot. While on the other direction, it's more of the other direction. So I think COVID did something very interesting, is it put you in a isolated place where it lets you ponder who you are. And what it does is after that, I kind of work on various startup, and I kind of experience this lonely chapter, which is one people not going to understand you, right?

People are going to think you're just kind of weird, you're kind of different and doesn't really. Yeah, just not like the rest of us. Right? Yeah. And they pass. And thanks for internet, thanks for the connectivity, you meet a lot of interesting people online. Although I'm anti-LinkedIn and social media, all those things, but I actually met quite a few interesting people on LinkedIn and quite a few interesting people on Twitter.

And the second thing is one, I don't think I'm one of those very smart kid where it's like, oh, they are winning the maths Olympics, they are winning the physics. I don't think I'm one of that, and I think give me a specific edge in just be very humble and okay to be stupid. I went to my first networking event when I was 16 years old, and all of them are 40, 50 years old, and I was in this suit that was way too tight for me because I haven't wear it since middle school graduation. And, I was on a growth spurt or something, and that was okay, right?

I was out of place, but it was okay and- I didn't have an ego about it, and I actually met my very first mentor, Vlad, over there. He used to run CoFounder Conduct, one of the biggest co-founder organizations in Seattle. And for the last three years, we've been calling once every two weeks on Sunday morning, including today. Whoa. I just called him today, this morning, and he's been guiding me through every step of the journey while the first time we met is I was at that place is because I was just going to him and be like, "Hi I'm 16, and I don't know what to do.

I'm interested in starting a business, but I have no idea." Just be really stupid, be really humble, and that's okay. You learn. Yeah. Definitely. That's actually a great experience of having a mentor that was so generous in giving you time and guiding you all the way through. And I think there is one argument that's behind your product is saying that you want human connections be more human in the sense that probably one.

I just want to get back to the idea, what does being more human mean to you? And also because there was another post that you share about some of the great founders who have been so committed and dedicated to their startup space that they give up other parts of being a human besides being a founder. Do these two aspects contradict with each other a little bit? Or just start with what being more human means to you.

Yeah. It's very philosophical, and it's a very personal question right? And for me, personally there's a few experiences I think back a lot on, and one of them is Bradley. He's the guy who I was talking about, my mentor who was graduated from USC. And he's the guy I met in San Francisco. And I went to this event this biotech event. It's called Biopunk.

And I'm not a biotech person. I was just there because I want to learn. And I met him over the event. We chatted once a little bit, but after the event I was like, "Let's stay in touch," right? "Let's exchange contact." But he was asking me, "Oh do you live in a hotel? Where do you live at?" And I would say, "Oh, it's in the, I think, Fisherman Wharf." And he was like, oh, he's actually going that way too, so why don't we just go together?

And so we end up just going on this very long walk at 10:00 p.m. It was cold. In San Francisco, you know how cold it gets. All right. And we're just talking for hours, talking about starting a company, but also spirituality. A little bit about his experience with starting, being successful, and shopping for Ferraris, but also bankrupting. And having a company that failed and restarting again.

And we were talking about all those different things. How's the vision of the future of biotech. And that walk is such a good experience to me because the whole walk, one, I was learning a lot. I think that's part of what being human is. You're always learning, you always grow. But the second one is I was very in present, and I still memorize that interesting experience.

And just being in present, being in real life, doing something with another human being, or a few human beings, and having a very deep, profound conversation where you are very vulnerable. And we all wear all sorts of masks and barriers while we are in school or going on a street or talk with people, but sometimes when you pull down those masks and when you're being very vulnerable and share what are the problems you have, that's very human to me. A very transparent, very honest, very earnest conversation. And I like to create more moments of that.

And compare that to what technology have done is. Do you remember what's the last 10 videos you scroll on Instagram? Nope. Not many people do, at least. I don't. Yeah. And I'm still staying in touch with him today. I think it was last Thursday, he went to USC for a biomedical event and we catch up, and we ate lunch in this sandwich shop that he told me his dad used to bring him when he was a kid.

And we talked for a while. He gave me advice on Humanest, and it's amazing. It's very great to really have the people you trust. You can be vulnerable, but also just go do things together. And yeah. It's being human. I think we're wrapping up and- Okay. No wonder why one of your core values is serendipity. Yeah. Okay, so wrapping up, just final question.

If I watched this episode when I was in high school, I feel like I would've started thinking about visualizing the many ideas that I had. But I'm not a high schooler anymore. But for the high schoolers that could be watching, the young founders, not high schoolers only, sorry. Young founders, what kind of advice would you give them, especially in face of challenges that they face as young people?

Yeah. I see adversity as very interesting and- One is don't complain. What I'm saying is right now, it's not like I'm saying something novel. What I say have been said before but why do people keep repeating that? The same things. People don't apply them. Like you heard, don't complain, but people still complain. So it's like I'm just going to share the same thing, just repeat the same thing, just because they actually work.

Those old things your mom, your grandma tell you, they actually works. Don't complain and don't whine, and adversity is good because guess what? Just like the hero's journey, the bigger the dragon the bigger the reward, right? Because you've been through adversity, you can share with your kids that, "Oh this is actually such a cool story." What kind of story is that if you're just born you are a millionaire or billionaire, or you just solve world hunger without actually doing anything?

And people love it because, well, you're just born and that's boring. Nobody wants to hear that kind of story. So yeah, adversity is great. The bigger the dragon, the bigger the rewards. And so it's great. By definition, it's great that if people don't believe in you. That's great if people don't like the idea. If people mock you, that's great. That's awesome.

If you don't know how to code, that's great. Learn how to code. Yeah, same thing. If you're scared of public talking, good. Now you have the chance to learn how to public talk. And yeah, just that. Just embrace adversity and even chase after it. Even trying to get onto it. And then second thing is similarly, but be stupid, be humble, be ignorant. Because once you know how ignorant you are, once you accept that your own lack of knowledge, that's where you can truly chase after knowledge, you can truly chase after mentors.

You cannot really teach someone who think they know everything. Just how you cannot wake someone up who's faking default, who's asleep. I don't know if there's a relationship with that. I'm just making that up. But yeah. I got one last more specific one followed down that line is, do you have any ideas to share on making pivots? Because I think you do make a lot of pivots.

Do you have any suggestion on when you should actually take the pivot or continue moving- down the previous direction? Do you have any ideas on suggestions on that question? Yeah. So before we even pivot, I would say one of the biggest lessons I learned is from Paul Graham, again, is you have to build it for yourself. Because when you are young and when you're just starting off especially a first-time founder you are so untrained by the status quo where you have a higher sensitivity on what the demand of the future is like.

You have the higher sensitivity on what are the problems with current status quo. And you have to build it for yourself and around the journey, you will come. And this is how it works, which is you start with a thesis, and you start with how you think it's going to look like, and you build it out and realize, "Ah actually, I don't really like it. It's not really building for myself." Because maybe some lack of perspective you didn't consider.

Like, "Oh, actually, I built this AR wearable, it's cool but when I wear it, I go to events, people are asking me, 'Oh, are you recording me?'" And I was like "What do I reply?" And they're like, "Oh, I don't want to be recorded, so I don't want to talk with you." Nobody ever told me that, but that's what people mean right? Yeah. That's like I solve a part of your problem. I thought I was building for myself, and it did solve part of my problem, which is having me follow up.

But once I build it, I realize I'm not building that for myself, then I pivot. So it's kind of running a lot experiment. And you have to be careful, which is not do mentally pivot. You actually need to do something rather than just changing concept because the next thing is very cool. But the moment that you realize, you feel that this is not it, this is not something you like, and it's not a small part.

It's not something you can improve. It's more like the way it's designed on a systemic level. That's the moment you just have to pivot, and there's no shame on that. Just start brand new and that's okay. That's great. You learn something. Great. I think that wraps up our episode perfectly. Thanks again for joining us, and if you want to know more about what Zion's building, check out his social media account, even though he's anti it in the past.

Also check his website humanest.ai and also follow us, which is 18-VC.com. Thank you so much and see you next time. Thank you.