Tao Zhang (Co-Founder and Chief Product Officer of Manus AI) joins Chappy Asel (Executive Director at The AI Collective) live on stage, just 24 hours after the launch of OpenAI's agents, to break down what makes Manus different, and why it might be even better.
This rare behind-the-scenes conversation delves into what it’s like to compete head-to-head with the biggest names in AI, exploring product philosophy, user experience, global strategy, and the future of work.
Transcript
Chappy (00:24:13): Come on up! We already met. But let`s have the round of applause for Tao anyways.
We just saw a direct head to head comparison of Open Ai. So the talk of the town, everybody was talking about Chat GPT Agent, right? Which just came out yesterday. I want to hear, I already seen that it performs better, I want to hear, I already seen that it performs better, that Mandus performs a lot better, but I want to know what was going on your Slack threads, but I want to know what was going on your Slack threads, like the moment it came out, right? What was the first message that got sent? What was the instant reaction? And how has that evolved over the last like 24, 48 hours?
Tao (01:01:42): You mean yesterday?
Chappy: Yes. Okay, yeah. Like walk us through when it actually came out, when you were looking at it, initial thoughts, then you're running the evals last night.
Tao: Yeah. So I think it's like this. Yesterday we saw maybe it's the open sourcing model because Sam is talking about the open sourcing model for Manus. So we saw maybe the release of the open sourcing model, but instead we get the check with the agent, which is like-- we think it's a very great product. And because we also have some friends in OpenAI, so we know there is something coming. But we don't expect that this will be a very similar thing to the general agent part. So the first thing we do is to go to the website to check all these use cases and all these demo videos. And after watching all these demos, we can be very sure that this is something very like all the generations in the past four months. So the first thing is we feel really like, I don't want to use this word, but actually we feel really happy about it. Because, you know, to be a startup, all the investors, all the media will say that, what do you do if there are some very big companies, there is Frontier, AI Labs do the same thing. They will describe this thing as a very scary thing. Like, oh, the giant will get you down and you will lose. But actually, if you have started one startup before, you will know when there are some big companies doing something like yours. Actually, you will feel really happy because you know you choose the right way, you know. So actually, yesterday, I think the whole team is very happy. We just sent our first tweet. It's like, welcome to the game, you know. A very friendly town. It's like we think it's really great to have some frontier AI labs to join this trade. So that's our first reaction, which is feel really happy about it. Yeah.
Chappy (03:25:51): Yeah. I love it. I love it. And that's how it should be. Yeah. Okay. So you are one of the first to enter the AI agents, generalized AI agents game, right? Yeah. You blew onto the scene. Everybody was talking about Manus, right? Everybody still was talking about Manus. Yeah. Now you have open AI agents, and you have a lot of people in general entering the scene, right? Everybody's going after the generalized AI agents. What makes Manus best? Put it quite simply. What is it? The technology? Is it the UX? We saw it was better, but what specifically?
Tao (03:54:23): Actually, we did a lot of things to make Manus the best agent, in my opinion. But if you let me choose one thing, I think that must be that sentence just written at the bottom of our official website. If you go to our homepage, you will see that sentence now. I personally put that there, which is less structure, more intelligence. Just like my keynote, like maybe one hour later, is like in the fundamental part of Manus, we choose to be the hand. We don't want to be another brain, you know. We just want to build a hand for these very smart models. So we choose to build the Manus in a very simple structure. But you know sometimes the simplest thing may be the hardest thing. So we choose the hard way. We choose to build this agent in a very simple structure. And we leave all the intelligent part to the models. But it turns out this should be the right way to build a general agent. Because once you just forget about all these predefined workflows, all these like to use different system prompt to limit the true potential of LLMs. When you figure out how to do this simple structure in a proper way, you release the true potential of LLMs. So I think if you need me to choose one thing that what makes Manus the best general agent is that sentence, is less structure, more intelligence.
Chappy (05:28:41): Yeah. Less structure, more intentions. Okay, yeah. So that's why it's better right now. I'm also curious to know what you think is your moat, right? What is your persistent edge? We can talk about, is it the team? Is it the culture? Is it your approach that you just said there? Is it go to market? Because this is going to be a many years long game, right? So how do you keep that persistent edge?
Tao (05:50:29): We answered this question in the past three months is to many, many investors. Everybody is asking about your mode. But actually, I think being a startup, sometimes you don't have all these resources. You don't have all these competing advantages for these big companies. So we never thought the mode is like some technology secrets. So actually, we share everything. Just one week after release of Manus, when I'm doing all these user meetups around the the world, we just share all the concept of how we made Manus, just like the less structure, more intelligence bar, and a lot of other sharing. And just one article we released this morning for our chief scientist, Pete. He is just explaining the whole context engineering philosophy behind the Manus on our official blog. We never hide these things. So we never think that the mode is based just on some technology secrets, you know. We think the only mode a startup can have is just to work really fast. Yeah, like, you know, I know like someone in some frontier AI lab say, Manas is six months ahead of other AI agent. I think it's really a accomplishment for us because we think we are only three months ahead. So, but we want to keep this advantage. It's like every time, you know, when others catch up with us, we're always three months ahead. That's the only way to build the mode to be as a startup.
Chappy (07:29:32): Yeah. Exactly. From all of us here in Silicon Valley, we know that, right? That is one of the most important advantages of being a startup, right? Is that you can move quickly, much more quickly than the incumbents can. Another common Silicon Valley thing. Who here has read Zero to One before? Peter Thiel's classic Zero to One. Okay, we've got a decent amount of hands going up. Okay, we've got a decent amount of hands going up. You've read it too. Great, great. Yeah, everyone's reading that. You start by finding your niche and owning it, right? Land and expand. That is the common Silicon Valley mantra. Find that niche and then expand from there. Yeah. It seems to me, and you can tell me if I'm wrong, that you are taking a very broad approach. We're going to solve everything. We're going to do generalized agents and win the whole thing. So talk me through that strategy.
Tao (08:11:51): I think it's really interesting that after Readiness Manus, when I came to Silicon Valley in March, I talked to a lot of agent staff, agent staff founders and agent researchers, and what I found out in Silicon Valley is that everyone is building vertical agents. Yeah, it's like a sales agent, marketing agent, but actually it looks like no one is building general things. I think that's our opportunity actually. That's why we get so much attention. It's because we may be the only one, maybe the first one at that time to build some very generic agents for average people. At that time, you know, like around the world, everyone is building coding agents. We have a lot of coding agents in the past six months, right? But for us, it's like we just want to democratize the AI power to normal people. Yeah, we don't want to just build another thing for ourselves, for our friends, for our colleagues, because we are all living in this tech industry. We think that AI technology is mature enough, and we want to democratize the AI power to normal people. So that's how we come up with this idea in the first place, is we want to build something for normal people. And when you want to build something for normal people, for average users, it seems not right to just build a very vertical product. Why is that? Because in the past two years, I was in many AI hack zones. I was a judge. And in all these AI hack zones, there will always be one type of ideas coming out, which is AI trip planner. In every AI hack zone, there will be an AI trip planner. Every time I saw that, I was like, oh no, not again. But why is that? It's not just because about the technology will not work. Actually, technology works. Because in these AI hack zones, Trip Planner may be the best project. Because it looks fancy. It looks awesome. But why I don't like it is because this is the problem from the business side, from the business model side. Because when you are targeting consumer market, the frequency really matters. As normal people, they only travel like one times or two times a year, and some people, they never travel at all. So if you want people to remember, like you are AI trip planner agent, you want people to remember you when they need you, which means you have to put a lot of money on advertisement to let people to remember you, or when they search something, they will saw your ads and they will get to you. So frequency really matters. So I think it's really like the B2B verticals. In B2B, you can just do verticals, and you can get some customers. But in the consumer market, if you go too vertical, you know, you go too long tail, it's really hard to get these users.
Chappy (11:21:40): So you're begging me to ask another question. And that follow-up question is, why B2C versus B2B? It's another common Silicon Valley mantra. It's another common Silicon Valley mantra that it's much easier to do B2B. So why B2C? And you can keep this one a bit shorter, because it's related to what you just said.
Tao (11:35:28): Yeah, of course. Yeah. I think there are a lot of reasons. But I think maybe the true reason is that in my whole 15 years career, I spent more than 11 years in consumer product. So it's very easy for me and our CEO. We come up with the idea that we should build something for consumers.
Chappy (12:00:07): That's a good answer. And you're nailing it. You're absolutely nailing it. Okay. Another thing that fascinated me that you said right near the beginning of your keynote. Is that you're not building a foundation model. Is that you're not building a foundation model. Right. I'm interested to know you have, let's say OpenAI now is your main competitor. I know it's been 24 hours, but let's say they're your main competitor. They're fully vertically integrated. Right. And maybe you could say Cloud Code is another.I use Cloud Code every single day. I love Cloud Code. They're fully vertically integrated. Right. And you see how like Cursor now is struggling a little more because they're not. So, but it sounds like it's a very intentional choice to not be fully vertically integrated, not building your own foundation models as well as your own agents. Walk me through that as well.
Tao (12:38:25): Okay, so I just want to clarify on this question is about to not being a model company, how to compete with them, is that the question? Yes. Okay, cool, yeah. So I think in this year, 2025, I think maybe this is the best year for AI application layer company, you know, to get started. Why is that? Because I think maybe three years ago, model, the top tier model, is a very rare thing. At that time, the only model we can use is open-ended model, right? But in this year, we have a lot of very advanced models from many frontier AI labs. So I think the frontier models are getting commoditized. So it's like you can just choose any advanced model and at a proper price. So I think to be an application leader company, the advantage is that in any given moment, you can always choose the best model, you know, for the right task. So that is kind of like a advantage. So you can iterate fast Because you don't need another three or six months is on model fine tuning, things like that. You can just use the existing best model. So I think this is really the best time for AI startups doing applications.
Chappy (14:08:56): Yeah. It makes a lot of sense. It's like how nobody does their own cloud infrastructure anymore, right? Everybody uses AWS or something like that. Right? It's commoditized. Makes sense. How about…You're focusing a lot on the user experience, right? For the everyday person you're traveling all around the world, talking to users, understanding use cases, yeah. Talk about how user experience, user research goes into how you're building your product?
Tao (14:33:49): Yeah yeah so at first we are generating. So we're already happy you know when we build Manus we have a lot of moments was like what the hell it's like how can he how can i do this you know yeah. we have a lot of moments like that so when we release manners we just want all the users to have their own aha moment. But after two months, we found that for most aha moments, they're coming from very long tail, because they can't find a solution ready on the internet for these very long tail tasks. But when we're doing some user interviews with our users, we found out maybe they will find their first aha moment with these long tail use cases. But at last, they can't do that on a daily basis. They still have to solve their daily work. So that's kind of like some inspiration we get from all these user interviews. So right now, we're kind of more focusing on knowledge workers. Yeah, it's like how to be their best research tool, how to be their best documents co-pilot tool. It's like when you want to deliver a PowerPoint, we want to be the best co-pilot there. When you want to figure something out, we want to be the best research tool there. So I think that's the learnings we never found. It's not just a focus on these magic moments. It's also about the frequency, the retention thing. Yeah.
Chappy (15:58:17): I love it. All right, I'm going to switch topics a little bit. And I want to talk about a little bit in the global politics area. I want to understand you are one of the most globally prominent companies internationally. And you are from China, I know you're now in Singapore, right? So I just want to understand, I think a lot of us here want to understand, what are the headwinds that you're feeling? If any. You can also say there's no headwinds. But I want to understand the headwinds that you feel being a Chinese or Asian in general company.
Tao (16:35:24): OK. I think it's just like there are a lot of talents in Asia. Because we know, actually we have a lot of engineers and they are very good, you know, and we work really hard. And in this AI train, we think, you know, US must have the, you know, the best models out there and the most mature, like, infra layers there. And for this company, like Butterfly Effect, from the first day we started this company, we targeted to the global market. So on the first day, even for our previous, product like Monica, which is a Chrome browser extension. We don't provide services back in China, because at that time, for Monica and now for Manus, we only use like Anthropica's model and Google Gemini's model. And in their policy, they don't allow us to provide service in some specific countries. So this company, on day one, is focusing on delivering the best AI user experience for the global market. So that's kind of like the mission. And that's why we choose to relocate to Singapore. That's also one reason.
Chappy (17:47:38): Yeah. Well, I've definitely noticed the global first approach. That definitely shines through in everything you do there. I'd also be fascinated to know, what does a user using Manus look like in different parts of the world? Because I think a lot of people building here, myself included, think about US audiences only. So I'd love to hear how you think about that.
Tao (18:04:43): Yeah. That's a very interesting one. Because I think for the past two months, We have some user studies across the world for different countries. And we found out that actually there are some very big difference, very big difference. I can just give you two example, if that's OK? Two example. Yeah. So it's like we have a lot of users coming from Brazil. Yeah. And in Brazil, we found that most of the users, they don't use computer to access Manus. Like maybe 90% of them are using their phone to access Manus, which is really different when you're comparing the users in US. And also, the background of the users are very different. I think in the US, it's more for knowledge workers. But in Brazil, I think most of them are small business, small media business owners. So they will use Manus to generate some product page or service page. They can show it to their customers. And they don't care about the domain name, which is the manus.space. They don't care about it. They just want web pages. And they can send it to their customers. So the use case is really different when comparing to different countries. And another thing I can share is that in June, we released a new feature with the Manus Slides. It's like Manus can deliver very beautiful slides for you. And when we look at the Manus Slides features, like how different country users use it differently, we found out that because we are not just generated the slides for you, we also have an edit feature, which is like you can edit the text, the image, after we generate it. Because now, AI can't be perfect. Sometimes you have to do some final touches for your slides, right? And when we look at the adoption rate of that feature, edit feature, we found out that on a global average, it's 20%, which is like 20% of users will edit their slides after we generate it. But in Japan, this number is 40%. Yeah, they're super cautious. They take very seriously about the size. You know? So they have a very interesting, like, different behaviors across different countries.
Chappy (20:24:03): All right, awesome. Let's switch to talking about the future. Future plans, what you're doing. But also, I want to hear from you. I think one of the awesome things that we get being here in our world is we think a lot about the future. We spend a lot of time thinking about how, I hope, spending a lot of time thinking about how the technology that we are building impacts the rest of society. Certainly with AI Collective, that is one of our big focuses. Talk me through what you think the future of work looks like. Because you are so deep in this, and you are building the future. What does that future of work look like, and what part does Manus play in that?
Tao (20:56:39): Yeah, I think talking about the future, because I'm the Chief Product Officer, I just will think this question from the product perspective. The first thing is about how to run a task. Actually, right now, you still need the human to trigger Manus to run a task. You have to assign a task to us. So I think in the future, there are many ways to trigger the task. Yeah, it's like right now we have a scheduled task, which it will automatically run by itself on a daily or weekly basis. And in the coming weeks, we will have some new features that Manus will just split the main task into many subtasks. And all these tasks you may not get aware of, but Manus will do a lot of job behind the scene. And also, I think in the far future, maybe next year, after you get more familiar with the agents and you have more trust into it, maybe with your personal context. So maybe Manus will work by its own to prepare something before you even ask. So that is kind of like this future. It's like a real personal assistant that you trusted. That is one way. Another way is like, actually, I think agency is always about tool use, use tools. Right now we have a virtual Linux machine, right? I think in a not very far future, Manus can also have a virtual Windows, virtual Android, so which can, Manus can use many more apps, and also many more services. So maybe Manus can prepare some event or emails for you, things like that. So that's kind of like the future, yeah.
Chappy (22:47:14): Okay, yeah, so that's the future of the product, and I love it, it's a very exciting roadmap. So extrapolate this a bit for me, right? Put on your futurist hat, you're allowed to speculate a little. So let's say that Manus is going to become the next Google, right? Manus becomes a multi-trillion dollar company. This is also in the future. I mean, that's certain. Okay, now let's talk about the uncertainty. You have billions of people using Manus every single day. And they're automating large swaths of their day-to-day tasks, work, life, and otherwise. What does that future look like for society, for our economy more broadly? And this is speculation. This is you speaking as a person and not as Manus. Don't worry about that.
Tao (23:31:54): Yeah, yeah. I think it's just like after we human invented cars, we can get somewhere we never think we can get, right? Because only with cars we can get as that far. And I think with Manus, with all these agents out there, we think agent market is really big. It can contain a lot of different type of agents. But I really believe in this agentic future, which is that all these agents will just expand the boundary of humanity's productivity, which means one man, one woman can deliver much more than we can deliver now. So which means at that time, right now we still have some, in the whole world, we maybe have like a top 500 companies, and each company maybe have like thousands of employees, right? But in the future, maybe there's going to be like top 5,000 or top 50,000 companies, and maybe each of them may not have that employees. But this does not mean you will lose your job, which means we can find some new things to do, some new business model, some new services. Just like we invented cars, then we will have some jobs for gas station, for self-driving tours, things like that. So that's kind of the future we are looking at.
Chappy (25:02:27): You're democratizing AI, right? I love it. Okay, so last question, and then we'll turn to Q&A. So start preparing your questions and everything like that. I was re-watching the Manus announcement video that went super mega viral all over the world this morning. I was re-watching it. And I believe it was like the first or second sentence and I believe it was like the first or second sentence was that this is a possible glimpse into AGI. Now, you're not the first company to say that. There have been a lot of companies. GPT-4, I think, was a glimpse, at least before they edited the paper. It was a glimpse in AGI. How close are you trying to build AGI? Like all the big foundation model companies, is that like an explicit goal of yours? and regardless of that, what is your timeline to AGI? How near do you think you are? And then what role, if you're not building AGI yourself, what role does Manus play in that?
Tao (25:57:33): I can get your question. Actually, from a builder's perspective, I think we've already reached AGI many times. But the problem is that whenever we human reach AGI, we will always set a new standard for the AGI. It's like if you drag somebody from three years ago, drag him to now, and to let him to watch the magic of Google V3 video generation model, or open eyes image generation model. You'll get shocked. They will not think this is AGI. This is a magic future. And they will think maybe it's like 50 years later. But we've done this under less than three years. So that's the problem. The problem is not about whether we've already reached AGI or how far it is. It's like we think human is a very adaptive animal, which is like, you know, we are very kind of used to it. It's like whenever we find something magic, after one month, but in AI trends, it's after one week, we will get used to it. We say, "Oh, this is something I used from when I was growing up." Yeah, it seems like that. But I think the problem is not about whether we can reach AGI. The reason is always about how to make sure in the future, like maybe six months ago, one year or 10 years later, all the people, all the average people can get access to the most frontier technology. I think that's what we care about the most. Yeah.
Chappy (27:36:11): It's a noble mission. All right. Round of applause for Tao. All right, we are going to go into Q&A, and I'll try to make it relatively quick because I know it's Friday night and we all want to get to the, whatever we're doing after this. I also have five pairs of AI Collective socks. So the first five people to ask a question get a pair of socks. And you were straight for it. I am going to walk this mic over to you because I do want you to speak through the mic and you also get a pair of socks. Okay, yeah. Here's the first one. I remember, yeah.
Guest 1 (28:07:21): But that was my question. How do I get a sock? I'm kidding. I'm kidding. Hey, Tao, I love it. I've been using Manus just for a week. And the question is, unlike all the demos here, my usage of Mauos is very different. I'm building Chrome extensions. And right now, for the first time, I downloaded your Android app. And I asked it to build me an Android app. So it's doing this for 15 minutes. But your Chrome extensions are great. But let me ask you a question. With all of this focus on content delivery, like creating slides, creating PDF, creating Excel, you seem to have this entire feature of actually creating products. Why are we not talking about that focus?
Tao (28:52:44): I think that's just the limit of the current model's capability. Because about their context, the limit is not ready for a large project, which can work. Like your use case, you want to build an Android app, and normally that will be a very super big code base. And I think right now, I think the capabilities of these models can't handle that large projects. But we think just like our philosophy, with less structure, more intelligence, it's like because we keep the structure as simple as it is, so maybe like six months later, when the models evolves, maybe we can solve these problems, But just not now. Yeah. OK.
Chappie: Great. OK. We're going to try to make these questions a little more rapid fire, OK? Because I do want to give people time to come talk with the Manus team after and everything like that. So let's keep the… Tao, if you can keep the answers to like a 1:30, that'd be great. OK, you were the loudest. There you go.
Guest 2 (29:53:50): Thank you. Thank you very much, Tao. Love Manus. And I have a question related to MCP. So we are seeing a rapid rise of MCP, Model Context Protocol, proposed by Anthropic last November, which also enables LLMs to connect with external tools which also enables LLMs to connect with external tools through MCP servers. And in fact, I'm building an MCP server marketplace, mcpstore.co. There are more than 24,000 MCP servers. So do you see this…do you see the future… How do you see the future of MCP server ecosystem? And also, like for the future of Manus, Do you think it will be compatible with MCP servers? Or do you see it more like competitive?
Tao (30:39:22): So I think it's a really great question, because many people ask us, does Manus use MCP? But the problem is that when we started this project back in last October, there is no MCP. So we kind of build our own mechanism of MCP ourselves. Yeah, inside Manus, we have some very similar mechanism to that. And I think it's very easy to get compatible with MCP mechanism. And why we're not doing that right now is because right now I think the ecosystem is not ready yet. Because if you are really building an AI agent, because AI agent is all about function calling, about tool use. So if you put a very large MCP server in, which means maybe you will bring hundreds of different functions to be caught. And right now, because of the limit of the current model's capability, if you give them a tool set more than maybe 30 or 40, they will have some very significant performance reduction. So I think that's a problem. That's the model company will solve from their side and also the MCP ecosystem will also like figure out a better way to do that. And also another problem is that right now because the MCP ecosystem is kind of like all from third-party volunteers so actually each MCP servers quality is is not guaranteed and when you are doing function calling it's always about how to describe the function in a very proper way and a very precise way. But if we just get into some unknown MCP server and we don't know how they describe their actions, their functions, we can't guarantee our user experience. So that's why we are kind of evaluating the ecosystem, but may not get into the ecosystem in the near future.
Guest 3 (33:00:51): So let's go specifically to the user's manners and the user's behaviors. So when we throw a task to manners, and it will get started and then will show how it works, and they will indicate, you can interrupt me at any time during my work to provide new information, or just a plan. But my curiosity is that, how many of the users will really interrupt the AI and just some kind of new plan? You just mentioned the Japanese people, the Japanese users, probably 40% of them will re-edit or revise when one task getting accomplished. But how many people will really interrupt, right? Interrupt, because as a carbon-based intelligence, I always feel I have no brave. I'm not brave enough to interrupt the AI to do something. But I'm brave enough to revise something, or to get something, just adding something, just when he gets one task accomplished. So how about that? Just a question.
Tao (34:08:22): I really love this question. Why is that? You know, before we release Manus, we put a lot, really a lot of engineering effort to make Manus can be interrupted. You know, we put tons of work into making sure that. Because you know, actually I think this may be the first consumer experience that you can interrupt the AI process. Because before this, all you can use is chatbot. Chatbot is like you ask a question, and you will wait until the AI finishes its answer. And then you can ask the next question, right? But in Manus we put a lot of effort into that you can interrupt any time. But unfortunately, just like you described, only a very small percent of users…That's true. Very small percent of users know or they are there to interrupt the process. Just one thing, unfortunately. We want people to interrupt, but it seems that they dare not. Yeah.
Guest 4 (35:12:46): Hi, Tao, and thank you. So thank you so much for hosting this and showing us different use cases. Very appreciate the talk. So I'm asking from a user's perspective. So I saw the OpenAI agent and also the Mauas agent. I think it's super cool. As a pro user for both, I'm curious, what are the different use cases between Manus AI and OpenAI? What are they good at and what are they not good at, just from a user perspective? So I know what kind of task I can go for.
Tao (35:47:31): Okay, so I think if you are both pro users, I think our advantage will be just we focus on the deliverables. Yeah, we are not, you know, we are just like a sound chat bar you can just discuss with. We're just focusing on the deliverables. So whenever you don't want to put a lot of time into a task and you just want the final thing, like a very organized and detailed report or like a very good layout slides. Yes, when you want these deliverables, and you can just come to us. And also another thing is like, maybe you don't have a lot of time, working back and forth to a lot of following up questions. In Manus, like for me, I use Manus just like this. Every time it's like, okay, just assign a task. And I will never watch it again, and I will put it back to my pocket. Because I know, like five minutes or 10 minutes later, I will get the result. Yeah, so which really give me confidence, you know, yeah. So that's the thing here.
Guest 5 (36:51:28): Hello, everybody. So I come from Tunisia, North Africa. My question is about empowering African youth. So there has been great cooperation between China and Africa. And the Chinese government is empowering African youth. And African youth is contributing to the development of China by skilled knowledge transfer so i'm asking you if you envision any strategy to empower african youth by knowledge transfer and encouraging african youth to embody AI development in Africa and AI education, of course? Thank you.
Tao (37:36:12): Okay yeah i think you know that may be like too big a casket for us because we are just a very small startup and we struggle to survive. But definitely, I think in the future, if we survive through this very intense competition, we really want to give more power to all the youth around the world, not just in Africa, just everyone in the world. Like three months ago, we started a campus program, which is like we will offer some free credits, free accounts for different universities. That is also the reason why we start these programs. It's like we want all these young people to have a chance to test the newest technology. Yeah, so that's it.
Chappy: All right. Round of applause for Tao. Big round of applause. All right, I did want to save 15 minutes so you could come up and talk to any of the Manus team. They're all bunched up over there. Thank you all so much for coming tonight. Um, that's all we've got here.
Comments