Building AI for the "Third Nation" - 6M Users, $10M ARR from Singapore | Bruce Yang, Agnes AI

Speaker A: You don't have to do a lot of detailed process work because AI will handle it. We're not going to develop software for human anymore. It's mostly software for AI, which is the skills and AI will handle all the hard and dirty work. That's why you see, um, a lot of SaaS services do have the mindset that we need to provide the most convenient product for human. But the problem is humans are not going to operate on the software anymore. So uh, a lot of uh, SaaS databases are not going to exist.

Speaker B: Hello everyone and welcome TO uh, episode 88 of Colors of Web 3 and Entrepreneurship. My name is Lum, your host. Uh, for those of you who are new to the Show, Colors of Web 3 and Entrepreneurship is a show highlighting the journeys of builders and innovators in Web three and entrepreneurship. Uh, some topics especially in um, AI and Web three make it technical, but we do our best to keep it accessible for most of you. Uh, here I'm today I'm joined by um, Bruce Yang. He's the co founder and CEO of Agnes AI. Uh, Bruce, can you give a quick introduction about yourself?

Speaker A: Yeah, thank you very much. Nam. So uh, I'm a founder and CEO of Agnes AI. I'm not sure whether we are that much into web 3.0, but we're definitely AI. So uh, we uh, are um, um homebrewed AI company from Singapore. We are uh, right now over 6 million registered users. Among m all the apps we have our own model, um, a group of cluster of uh, AI models supporting us, doing large language model, doing text, ah, image video models, doing the multimodal, uh, generation and a lot of cool things. So the one thing which uh, sets us apart from all our other potential competitors is we focus a lot in the Southeast Asia regions. Uh, a lot of um, AI companies focus a lot on the western world, um, America and uh, maybe paid market. We have a big vision of uh, AI inclusion or AI parity. So that we hope that uh, we're able to get 99% of the um, population who are not that much into AI, uh because of the paywall. Um, we get them to access to the coolest part of AI. That's part of our vision.

Speaker B: Uh, nice. Awesome. Yeah. So I think many people uh, watching the podcast would probably be curious to know what you did, how you came to this. Maybe if you can uh, deep dive and share with us, walk us forward from pick a point in time, maybe from your first job after college is okay, and then forward up to the present day.

Speaker A: Yeah, sounds good. So I, I, so I came to Singapore at age of 15 as a scholar from China uh, under, under a scholarship. So I, I've been uh studying as in Singapore rather um the entire high school. So after that I, I got opportunity to study abroad at uh, uh America uh at a center of the Silicon Valley which is UC Berkeley. So I have my honest undergraduate in math and computer science um and that was like in year of um late of 2000. So it's like I graduated when I graduated it's 2010 so quite some time ago. Um but that's the exact time of web 2.0 which we see a lot of cool companies built uh as that state and a lot of people like our professors, our um classmates uh go to the Valley. I mean we're already at the Valley go to some big companies or start their own uh ones and be something very cool. So I got the chance to also uh join Microsoft and then LinkedIn after I graduate and there was time which LinkedIn just went IPO and during my time staying at AH LinkedIn every quarter we be the business with the Wall street expeditions. So um, we're able to the quarterly cost we keep having our um price goes up during the time. So I was among the first to uh build a subscription for LinkedIn. That was time when LinkedIn started thinking about monetization and um, I'm one of the person in charge of the entire subscription team. Um so that's one of my journeys in the Valley. After that I started to build my own company. I had an early startup which focused on social media. Ten years ago in the Valley which is similar uh, kind of uh, um maybe a Snapchat, um that was. You see a lot of cool new social apps coming up and I was also m get featured by cnn uh Fox News. As a Chinese entrepreneur in the region you see a lot of uh entrepreneurs, Asian entrepreneurs right now. But back back in 10, 10 years ago it's not, not that much. It's still a rare scenario. Yeah so I, I, after that I, I been uh keeping my journey in my startup arena. I have been working uh in enterprise business in China also you know uh, um work together with my team to uh try all kinds of different things, get funding, sometimes succeed and sometimes not. And during the COVID I came back to Singapore to pursue my PhD. So I'm a native PhD, UM student. But the good thing is it's a time when you see the boom of uh AI especially with the introduction of uh Chat ChatGPT, uh 3 and we were able to see a huge amount of imagination spurred by ChatGPT and definitely together with my PhD, um, you know, fellow classmates in AI we found something extremely interesting which is building um, you know, the next generation or maybe a younger version of uh, chatgpt in the region which is Southeast Asia. This uh, is time in the late of 2020, 2024. In the end of 2024 we started this journey of Agnes AI. Yeah, Agnes AI does move extremely fast. So during the last entire year we see a huge amount of growth. Um, actually we launched our app in uh, July last year. Uh, I mean launching on the web in July, uh, last year and the app is introduced in September. So within about half a year from September we are already at 6 million resume users and our DAU goes up to 300 and we also reached the AR of uh, 10 million. Um, um, and you might ask how we compete with big players like ChatGPT or maybe uh, uh, I think one of the things which we optimize a lot is our model uh, efficiencies. So we build our upholstering a lot of our smaller models and using a genius workflow, uh to connect all them together so that we're able to uh, cut the cost to about 110 of that of most of the players in the, in the market. Um, that, that allows us to build a viable business in, in the region like Southeast Asia and could potentially extend to other markets like Latam, Middle east because these are the markets which are not perceived as well paid market and you have to optimize a lot of things at your side. That's how we uh, tie together with our vision of AI inclusion for sure.

Speaker B: Yeah. Awesome. Well that's very impressive career path and career, you know, professional that you walk us through there. Uh, so it sounds like you were living in the Valley in the US right after college for quite some time.

Speaker A: Right.

Speaker B: And then could you share a bit more? Like you said you were building a social uh, media startup right in the Valley. Yeah. What was that about? Share with us a bit more.

Speaker A: So that was like in the year of 2014, uh, where I just uh, uh quit my job at um, Linting. Um, you know a lot of people at my time decided to start a um, company uh in the uh, early of their 20s. So I'm doing the same thing at the time I uh, decided to uh, build a startup um, together with some of my classmates at Berkeley. We run a place in M. Emeryview which is right next to Berkeley. That was time. We see a lot of Things happening with introduction of Instagram, um, Snapchat. Ah, we just thought that maybe we can do something similar. We call it an anti Facebook app. We call it Sober. The name of the app is called. It's the anti Facebook app. Yeah. So we want people to uh, um, have all their posts to uh, auto delete it in 24 hours so you can post whatever you want but within 24 hours it's gone. It solves the problems of uh, the mental barrier of uh, keeping everything relevant. Uh, you have to uh, purge your content once for a while.

Speaker B: It sounds similar to like Snapchat, right?

Speaker A: Similar to Snapchat. But you know we're learning that Snapchat introduced the idea of um, 24 hour story and soon after we introduced this feature Snapchat and Instagram uh launch similar features as a card story. But this is a very exciting time and despite it's not a very fruitful journey because we decided not to continue after the second year because of lack of funding. We still have the sense of running our own startup. Um, we as peak of our performance we are reaching um total uh Register user of 1,000,000 and DAU close to without any payment. But it still give me a sense of how the market looks like because Agnes right now despite we have the models as a model company we also running very strong on consumer. So I think um, the startup 10 years ago definitely um, have sowed the seed in my heart to have a dream to build something huge for consumer facing.

Speaker B: Nice. Awesome. Yeah. So I'm sure, I mean that gave you a lot of fire and inspiration right back in the day. But yeah, I think that's quite impressive. Yeah, it's quite impressive as well. But I guess at that time probably had to I don't know fundraising because it was quite probably some VC was seeing as like a crowded space. Right. Or something.

Speaker A: It's a very crowded space. We did have some fundraising. We have $1.2 million million dollars um, as our seed fund. But you know uh, in a consumer app, especially at the year of 2014 with so many things coming up like TikTok going and a lot of big things in China US come up uh, one minute is like spend everything in like a couple months. So we probably can't survive in. That's why you know for this round of funding, I mean for this new startup we start everything with a much better perspective of how we should run the business. And we just goes around a funding of 10 million and we're thinking about you know close another round of funding of 10 million by end of March, end of this month and by mid of this month we're going to raise another uh 30 to 50 million. So in the game of consumer and AI, speed is essence. So we definitely do not really want to um, make the same um, risk make the same mistake of uh, not having enough fund to sustain our growth.

Speaker B: Yeah, that makes sense. I feel like we are living in an accelerated age or accelerated era.

Speaker A: Right.

Speaker B: Everything moving, speed of breakneck. So I mean I think for a startup to compete with big uh company speed is pretty much everything. So you do have to move faster otherwise.

Speaker A: But you see that this is only time where startups can compete with the big companies because of all the disruptive um, scenarios disrupting landscape, um. Bravo AI. So if three to five years later M when the entire landscape stabilizes, uh, I would say we don't really have a chance to compete with companies uh, with bigger resources. So right now time is essence. We need to be extremely fast. We need to risk money fast. We need to uh, uh spending money fast. We need to get our metrics up fast so that it becomes a positive cycle.

Speaker B: Yeah, for sure. Time is a very important essence. Cool. Uh, just one fun question before we deep dive into your current main company. Yeah. Can you share with us what do you do for fun in your free time or your hobbies?

Speaker A: I play ping pong. I'm a, you know a relatively good um, uh amateur player in, in table tennis. Uh and I enjoy not only playing ping pong but watching all the ping pong games. I was at the WTT championship in Singapore as a, as audience on the first, first role because I have some uh, good partners, business partners who provide tickets for us and my two daughters uh, are both um, Timothy tennis players. They are still younger like primary one, primary four. But I keep hope that they can become a much better player than me and uh, together as a family. We are uh ping pong fans. So this is awesome. I keep myself um, not that much of uh frustrated uh, not that much of you know. Uh. Yeah to defeat the anxiety with p. Nice.

Speaker B: Awesome. So do you grow up? Did you grow up uh playing ping pong already or did you learn when you were an adult?

Speaker A: Actually no, I had some acquaintance of ping pong when I was young but it's during the COVID which I'm like lobbying at home, couldn't go out, buy something to uh really uh have some time to uh, sit down together with myself to think about the future. I decided to grab a lifelong habit, lifelong hobbies of uh, ping pong. Because you can. This is a game which you can play up until year of 80s and it's a game which you don't require a lot of facilities. It's not like skiing or golf. You have to find people and you have to a lot of preparation. You can start ping pong anytime you like. So I have a ping pong table right next to my um, you know, office. Uh, I would play anytime I like.

Speaker B: Yeah, yeah. I think ping pong is pretty relatively, I would say like very simple to play, but actually really hard to get good. Right. It's one of those games I feel like it's play ping pong. Okay.

Speaker A: One of the game which you can keep improving until um, very long time. And especially if you are some of the young players, if you miss the age, you can't really be the best players. So it's a game which you won't really, uh, have a problem of finding opponents.

Speaker B: Yeah, for sure. Because I think we both know that in a lot of professional sports, I think the older you get, obviously your ability and skills, everything you know, is going to reduce. Right. Whether it's like maybe soccer, football, basketball. Right. Obviously you can keep playing good, like professional level forever. But ping pong, I feel like, yeah, the age will, I guess it still matter, but it matter much less. Right? Much less like the other, much less alternative. So like you have a somewhat of an equal playing field.

Speaker A: Yeah, exactly.

Speaker B: Yeah.

Speaker A: That's why the agents can be the westerners in the Olympic Games.

Speaker B: Uh, makes sense. Yeah. I think it's just like, I don't know, there's a book already like 10,000 hours, right. You practice 10,000 hours of something, you become an expert.

Speaker A: Yeah, expert.

Speaker B: I'm sure to apply for sports and Big Pong as well.

Speaker A: Exactly, exactly.

Speaker B: Nice. Awesome. Cool. Yeah. So, uh, without further ado, I think let's go ahead and maybe switch topic and deep dive into your current company, Agnes AI. I think it'd be good to maybe let's talk a bit about the background, like the founding story of the company. Right. Maybe deep dive a bit more into that and then afterward maybe you can share a bit of a demo, like product demo for our audience to see.

Speaker A: So, uh, as I mentioned. So, uh, before we start this uh, journey of agnostic AI, I was doing my PhD at NUS. It's a time where I really want to um, um, sit down, think about the future. Um, the second curve, uh, especially during the COVID there's not much I can do, uh, even for my other Startups during COVID Um it's very difficult to grow the company at the time. So uh, um I spent some time, talked to my advisors and decided to uh work on the area about you know at that time it's not really um too much about chat AI but it's still part of the big data. So very soon after the introduction of ChatGPT we decided to uh do a quick pivot not only in my uh PhD course but also my startup course. So we decided to focus a lot into the application level about the AI and AI agents. We are among the first to know about the um impact of our AI agents. That was way before the introduction of Meta, not to mention uh open cloud. So in about um 2024, mid to late 2024 we're already thinking about how we should build something huge with uh AI agent. And um, that was a time which we thought that is a biggest opportunity after the introduction of models because very few players uh can compete uh with the big players out ah there. But AI agents offered an opportunity. At the beginning of the journey we spent a lot of time thinking about uh doing something vertical and application level. But very soon we realized that it's not the faster way of growth. Especially with the introduction of uh Manus, you see that uh the entire general purpose of agent just automate a lot of things before everybody else. We soon realized that the most important part about the vertical agent is agent itself, not about the vertical bot. And this kind of things has validated by uh instruction of cloud cowork and skills. Cloud cowork is like um, a general purpose agent but with any kind of combination of SKUs. It can soon become vertical SaaS or vertical AI agent. So that was early one year after we have the same kind of uh, uh inspirations. That's why we decided to focus a lot into the fundamentals which is the general purpose agent. And we start to think about building our own models. Um because um there was also opportunity of seeing uh the um huge impact of uh deep seq um in the early of 2024 I think uh early of 2025 um which we say that we don't have to build everything from scratch. What it can do is focus a lot on post training and um, post training proved to be extremely um, extremely effective and with much less cost in terms of uh training and much less cost in terms of inferences because we're using much smaller models, uh much smaller parameters. Um so um, together with the kind of infrastructure which we assembled with um post training um Open source models, especially on smaller models build an um AI agent, um architecture which have the routing mechanisms as an infrastructure. We're able to build something um 95% um that of Soda models are even better sometimes up to um 98 or even surpass the Soda models with a cost which is like 1/10 of that of uh um the big models. So that gives us a lot of opportunity to think about how we can excel in the application level. This is only our first app. We're already in the plan of launching two other apps um this month. One is focusing on the video and image generation. You see the. The outburst of CDance 2.0. A lot of people are thinking about how this can transform the entire industry of movie uh and stuff. We're providing a similar model with uh relevant apps just recently and we are launching another app which is um enhanced version of um character AI which is compelling purpose but it has a modern model. The image and videos generated within the scenes so you can be your own director, your own actor in the things. It's an open ended story. So this is going to be two new apps um besides all in one Agnes. We're launching a third one, I mean a fourth one um um which is focused on app and web development. You will think that as um a similar version um of replit a lot. I will focus a lot on the mobile part. So we have a cluster of apps um supported by a ah cluster of models and each of the model is very cost efficient in terms of parameters um but very effective uh in terms of the capabilities matching up with sota uh models. This is the entire architecture of our current story.

Speaker B: I see. Cool. Yeah. So would you say that most of your model like open source right now

Speaker A: um we build on top of open source models. We call them pre trained models. But these pre trained models if you just do starting using the models I want to give you something like uh uh mediocre scores like 60 um out of 100 if you count the 100 as the best SOTA uh cross source models.

Speaker B: Yeah.

Speaker A: So we have to do the work to improve from the 60 to 90 or 95. So this is a part of which we have all the innovations and we built on top of this all the apps which utilizing our own you know um self homebrew models. Self trained models.

Speaker B: I see. Makes sense. Yeah. I'm sure you have a lot of like the team people behind this doing the a lot research. Right.

Speaker A: We have about 40 researchers, 40 researchers and uh 80. Yes. So altogether we have 150 in the team and we're expanding to about 200 within about one or two months.

Speaker B: Awesome. Well that's quite a big team. Yeah, for sure. Expanding very fast too. Yeah, I think it'd be very interesting. You know you can show uh, share the screen, share a bit of a. Maybe a quick demo for people to see the capability of the Agnes AI model. Right. The main product right now.

Speaker A: Right, that's the one that do that.

Speaker B: Yes. Uh, I can see it now. Yeah, yeah, I can see Agnes AI.

Speaker A: Yeah. Uh, it looks slightly different on mobile and desktop. Uh, let me uh, go one by one. This is our uh, current mainstream, um, I mean uh, flagship app. But we have another two coming up this week. So by the time this video comes up, if you search for company name cpnci, you should see multiple apps on both on our website and on App store. So let me see, uh, let me research. Um. Um, maybe um. Let me just.

Speaker B: Insurance industry. About insurance industry. Insurance.

Speaker A: Okay.

Speaker B: Yeah, let's switch because actually that's one of. Yeah.

Speaker A: In Southeast Asia, in Singapore.

Speaker B: Yeah, Southeast Asia sounds good. Yeah, Southeast Asia because we both here. Yeah, makes sense. Do you have to switch a mode to researcher or you just do?

Speaker A: No, I don't have to. So uh, it understands the content and does a triage itself.

Speaker B: Okay.

Speaker A: And we are pretty fast. So normally if you do research outside uh, from other apps it might take a couple of minutes. For us it's like um. Within one minute it should get a result.

Speaker B: That's quite fast. Yeah.

Speaker A: And our token per second is very fast. And you see all the emission. Oh, it's in Chinese because uh. It might have a problem because I asked previous question in Chinese. But let me.

Speaker B: Okay, I see. Is this a new thread or existing thread by.

Speaker A: Sorry, let me just please provide everything.

Speaker B: Okay. Yeah, please translate. Right.

Speaker A: Yeah.

Speaker B: I guess because it remembers you as a user. Right. Obviously because I guess you.

Speaker A: Yeah. Sometimes I'm testing on Chinese languages. So that's why.

Speaker B: Makes sense. Yeah, it's good to test.

Speaker A: It has a memory of uh, um. Uh, the user. That's why uh remembers uh, your preferred language. Remembers uh, all your preferences. So you put up a report um, with all the links and sources. Sources and uh, trying to get uh, something out within. Um. The second time I asked it only take like five seconds because I've asked the same question before and it got it.

Speaker B: Yeah.

Speaker A: And what I can do from here is. I mean generate a deck based on the above info.

Speaker B: Okay.

Speaker A: What I can do here is I can use slides and I can Go for pro version. And Pro version takes a lot longer, but it's very good. I would say it's very professional.

Speaker B: I see.

Speaker A: We can wait.

Speaker B: Um, yeah. What do you say is like the main difference between the. What is it the fast and the pro? What are the main differences?

Speaker A: The first version is mostly optimized for speed, so the PowerPoint should be done within two minutes. Um, for the pro version takes a lot longer, maybe five to 10 minutes. It's still much faster than most of the other apps, but it's very professional. Yeah. I'll be able to show you the results, uh, in a few minutes for sure. With the results coming up, we can take a look at our, uh, mobile part. Mobile is slightly different from our, um. From all.

Speaker B: Is this also same or is this a separate app like you also chat or you have more models here for the mobile app?

Speaker A: Uh, it's the same account. It's the same account.

Speaker B: Same. Okay.

Speaker A: And on mobile it's available on both App Store and um, from Android Play. And um, on the first tab of work, this is mostly about. About what we're seeing on the PC right now. But what you would see is. Let me get back. Let me continue.

Speaker B: Nice. Oh, it's thinking.

Speaker A: Yeah, it's thinking. But you see that it already make the chapter.

Speaker B: I see that. That's cool.

Speaker A: Summaries. You have 11 pages. Pages. So uh, it's still in the process of generating. Let's wait for that. But, um, um, besides the work part, you can. This is like what you can see right now is um, the same as um, uh, on PC but it also has a multimodal part which you see all the video and image generation.

Speaker B: Nice.

Speaker A: I also have uh, some of the images generated by me and my friend. Um, this is me, um, getting a haircut.

Speaker B: Nice. Oh, wow.

Speaker A: And also you see, um. And this is the third tab is like, uh, um, you know, a character AI which you see.

Speaker B: Character AI.

Speaker A: Yeah, yeah. Different characters. I could play different games. Um, and uh, the group about us, how we differentiate from other players is we have the, um. We have the video and image generation so you can be after.

Speaker B: Interesting.

Speaker A: And uh.

Speaker B: Nice.

Speaker A: It has a story, it has a plot, so you can communicate. Chat with your uh, character and once for a while it show up the video and image and it has the communication.

Speaker B: Okay.

Speaker A: It has a voice so you can

Speaker B: make some selection as well. You chat with it.

Speaker A: Uh, yeah, you chat with your character and generate the video and I see. And it has a voice signalization.

Speaker B: Okay. M. Interesting. Okay.

Speaker A: Yeah. This is part of a story. So that's how we have the all in one. Within my app right now. Uh, we have both work and search and stuff which is uh, similar to the um, desktop version and we have the video image generation and we have the companion uh, uh so we're going to separate this part, um, this piece of image and video generation and the companion uh part into a virtual app. But uh, we still keep the ALICE as only one. Yeah, so this is the idea. We're still generating the slides.

Speaker B: Yeah, that makes sense. Okay, so you said you would separate the um, character AI all the character thing into like separate app.

Speaker A: Right, right, right, right.

Speaker B: That's okay.

Speaker A: Yeah, it's interesting.

Speaker B: And would you say that uh, which app or which model is like being. Is the most popular, like being used the most by your user? 6 million user. Yeah.

Speaker A: So uh, for the 6 million users we have right now it's about half of uh, maybe half of the users are ah focusing a lot on the productivity. The other half uh, is more on the entertain part. Entertain part includes both the image resolution and the companion part. The comparison part right now actually is extremely popular. Uh so I think um, the majority of the payment, especially the payment from um. The majority of the payment, especially payment from um, um. Indonesia market maybe uh, uh including all the entire Southeast Asia market together with the developed countries like US and Japan, they're spending a lot of money sending a lot of uh, um. The most benefit feature is on character part.

Speaker B: Nice.

Speaker A: The size are coming up.

Speaker B: Oh nice. Yeah.

Speaker A: Okay, let me do a demo about this size.

Speaker B: Yeah, yeah sure, go for it. Yeah, just to see the quality of it. Wow. Looks pretty professional. I can see that.

Speaker A: I mean. Yeah, yeah, that's wait for you a few last minute.

Speaker B: Gotcha, gotcha. Yeah.

Speaker A: Once it's done here we should be able to do a presentation.

Speaker B: Yeah, I think that's pretty cool. Yeah. And then um, so I guess like the, the cost of using each of these model would be different. Right. And I suppose uh, video and what is it? Video model probably the most expensive I imagine.

Speaker A: Um, yeah but we do we, we did have a lot of optimization. So uh, you would think that video models are very expensive but for us we are able to optimize to uh, read that every second. Might take um from outside maybe um, a couple cents, uh one or two cents maybe that's possible to optimize. Um and in terms of text it's much cheaper. Um because we are running on smaller models we do a lot of distillation with a Lot of uh, uh, what should I say? Do a lot of um, um, quantitization. That means we don't need a big uh, GPU to run our models.

Speaker B: Yeah. Okay.

Speaker A: This is ready.

Speaker B: Okay. The deck is ready. Nice.

Speaker A: Yeah. Deck is ready.

Speaker B: Nice.

Speaker A: I'm able to do that.

Speaker B: Cool. Wow. Very professional.

Speaker A: Yeah.

Speaker B: Executive market overview. Okay, nice. Market size. Growing. Yeah, I like that. Okay. Uh, grow driver. Okay, nice. Yeah. Makes sense. Yeah. I know some about the insurance industry as well, so LPNC emerging. Okay.

Speaker A: Yeah. So you might want to subscribe to our uh, slides here generally.

Speaker B: Yeah, yeah, I like it. I like it. No, it's good. I actually I gave it a try before the recording. I gave the slide because actually tomorrow I have a seminar to run so I don't have time make slides. You may as well try yours.

Speaker A: This is very interactive. Uh, it's actually interactive each of each of the card.

Speaker B: Cool. Oh yeah. Okay. I see. Which is good. Yeah.

Speaker A: And if you want to do the huge market, if you want to do a quick um, you know, uh, edit, you can go with advanced editing which all the tech.

Speaker B: Yeah.

Speaker A: And you can move things around.

Speaker B: Nice.

Speaker A: You can add image and videos. Um, you can insert stuff. You can uh, do all you know, uh, font changes, anything like.

Speaker B: Yeah. So you do all of this in the web app online. Right. You don't need to like open PowerPoint or whatever.

Speaker A: You don't you know, make it very easy our web. But if you're interested in download the PowerPoint, we also support the feature but that's for all subscribers.

Speaker B: Yeah, I see, I see. Makes sense. Yeah, yeah, all good.

Speaker A: It's one of the things which we are able to get people subscribe very fast.

Speaker B: Yeah, yeah, I bet. No, I've seen like some. What is it? There's ah, a one, uh, I think called Gamma. Have you seen Gamma? Gamma or something? They generate slides, right? I think that's mainly what they do I feel. Generate slides and. Yeah, that's awesome. I like the quality. It's really good.

Speaker A: Awesome.

Speaker B: Yeah. Ah, and also makes sense of what. Yeah. Okay. That's a really cool demo.

Speaker A: Yeah. We have different kind of other cool things like image and video generation is for slides. But if you also look at all this image is generated by our own um, models. So you can see from these inspirations you can click into it and change. Do a remix with a change that change on the prompts to generate um, your own uh, image uh, and videos. And beside that I think uh, we're also coming with a feature called AI Sheet which Is uh, um, something for you to understand about. Um, uh, some data, uh, do some uh, analysis. Uh, about. Yeah, this is the feature.

Speaker B: Okay. So like you can upload an Excel.

Speaker A: Yeah, you can upload Excel and you can do analysis for the data.

Speaker B: Nice. Interesting. Yeah, so I guess I would, uh. I mean there's. I'm pretty sure there's still a lot of jobs right now in the world that's still running on Excel.

Speaker A: Right?

Speaker B: Yeah. Like business analysts, for example. Data, uh, analysis.

Speaker A: Can we uh, do an analysis of insurance industry in South Asia? Let's see whether the. Will be able to work.

Speaker B: Sure. Yeah. Yeah. Interesting. Okay. So it's going to pull up some data from online, I guess, like and do a message. Okay, nice. What would you like to begin with?

Speaker A: Uh, smart data analysis, maybe.

Speaker B: Okay, sure. Yeah, yeah. Smart data. Interesting. So it looks like the model is already like. Remember a lot of preferences about you.

Speaker A: Yes. It has a memory. Okay. Maybe I ask them to give me uh, some AI sheet, uh, to try it.

Speaker B: Okay. Yeah, that's why we. Yeah, I think something existing might be easier, I guess. Yeah, I see. Oh yeah, it looks like it's uh. I'm just reading the, the text on there. It's saying that they upload your data file. Yeah, it's actually asking. Yeah. Okay, makes sense.

Speaker A: Awesome. To give me, um, some uh, AI sheets to look at. Yeah. We can talk about this later. We can do a demo.

Speaker B: Yeah, sure. Yeah, sure. Later. All good? All good. Yeah, no, I think that's cool. You have quite a lot of uh, feature, right. AI design. And then is that AI design like for cover or something? But yeah, you can quit. Quit the sharing.

Speaker A: Yeah, you can do, you can do AI design for a lot of things. By the way, our AI design also support Photoshop. Photoshop features, like you can separate the layers, you can edit on the text.

Speaker B: Okay.

Speaker A: Text, you can remove background. It's all supported. Nice features.

Speaker B: Yeah, that's very important for like a lot of designers.

Speaker A: Exactly.

Speaker B: I actually. Okay, that's cool. That's cool. I think that'd be helpful on my team. I want to check that out.

Speaker A: Yeah, try it out. Try it out anyway.

Speaker B: Yeah, Cool. So, um, would you say that. I mean, I remember that during the uh, what is the event where we met? Right. I remember you guys said like the AI inclusion, but also like building a company from Singapore. Right. It's like, to compete with like, what is your like strategy with your company? Is it like you're trying to target more developing markets or what is your strategy? What are the user segment. Just curious. Right.

Speaker A: So uh, if you look at.

Speaker B: Yeah, I would like to hear something.

Speaker A: If you look at the current um, you know the completion in the global market, it's mostly coming from US and China. And if you look at US and China, yes, both of them are very strong. I also have a team in China so we know that China tenants are very strong. China have a lot of good supplies in terms of um, the talents in terms of infrastructure. And we're just seeing that there's a strong need outside of US and China which form a third party to compete with the big two players. Otherwise the um entire power uh, um in terms of the things are controlled by the two big players. And if you look at the conflict with US and Iran uh in the Middle east, ah as I heard um, the US army is already utilizing a lot of AI in terms of the um operation in the future. Um, it's not about the weapon, it's about AI. It's about whether you have AI or not. We don't want, we have a vision of not excluding the um rest of world um in the development of AI and we definitely hope that Agnes can be one of the effort to have the third party outside of US and China um, to have their own models, have their own infrastructure. This is uh, our long term vision and if you look at, besides this long term vision, if you look at um, the current AI inclusion it's not doing extremely well. But um, we have about 5% of people uh in the world using maybe 10%. There's about 10% of people um, 10% of netizens in the world using um Gen AI every day which accounts for about 500 million to 700 million. Um, yeah that counts about um, um maybe 10% of netizens. And within this 10% of netizens, um for subscribers it's only about 5%. If you look at the, the um, I mean ChatGPT subscriber, um mostly from the most active user is only about 5%. So um, this is 10% times 5%. Only 0.5% of the uh global Netherlands are using the advanced features of uh, um the generative AI right now. And the rest of 99.5% is neglected. It's not well covered. This is a good opportunity for us because we believe in AI inclusion, we believe in AI parity. We believe that in terms of development it's best not to create inequality, create an unequal uneven um scenario where the uh, advanced um AI nations uh dominate the entire world. Just like how um the armies are using AI to uh, do a lot of operations. We want to try to reduce this kind of conflict by having equal development. Definitely the third parties uh, are indeed of this kind of AI sovereignty. Uh, that's why we believe that Agnes can play a huge role in this process. I also want to um, pay my tribute to all the entire open source community even without the open source community when we're going to start from scratch which takes a much longer time with open source, um, the iterations of open source models we're able to build from something like Midway NS Midway um 60% and we hope to reach that to about 90% and we wish our models and our services can cover um, concealer to the less covered regions so that somewhere in the future at least it's a three party game. Um, us, China and the rest of the world and we want to contribute to the third player. This is our long term vision.

Speaker B: Cool. Impressive. Yeah, I think we need that for sure. Right? Yeah. Because I feel like in whatever industry right. If you only have two players actually in economics we call it duopoly. Right. And as you know from I don't know economic is my major. So anyway back a long long time ago I was going to college but dual poly mean like they have so much power right. Over people. But now if you have more it's kind of like fair market, like fair competition and it's just like more choices for the user. So I think in the end I think it's win for the user anyway and also much more healthy in the market now. So that's a very um, impressive and lofty goal that I think you know, set out. Yeah. And I think open source really help a lot. Right. For sure. Because I feel like with a lot of the. I heard it's like what is it? I'm definitely not a researcher on AI but I know like what is it? A mixture of models. Like it's like one of the new innovations in gen AI right?

Speaker A: Mixture of models.

Speaker B: Yeah. Is that what you guys using right now too? What's some of your models?

Speaker A: Mixture of experts. This is one of the things which introduced by Deep Seq, I mean advanced by Deep C. It's part of our structure for the models because we have a lot in terms of reduce cost at the inference level. We also have that in our models but we also have um, a uh, cluster of dense models, smaller dense models which serve each of them survey experts and we use agentic routing to achieve similar kind of effect like moe.

Speaker B: Nice. Awesome. Yeah. So and I'm sure that you said like the 40 researcher. Right? They probably doing research on this in your team every day figure out how to optimize this even further. Yeah, for sure. Yeah. So it sounds like many things actually going pretty well in your company but is there any particular challenges that you guys facing on a daily basis?

Speaker A: We need to move extremely fast. As mentioned, it's a game uh, of speed because we don't want to uh, see a situation where our development is limited by the money or by the market. That's why we believe in um, a positive cycle of raising a lot of money, spending very fast, growing the metrics and ah, with money again. So this is a cycle which we're trying to catch up with all the other big players. But at the same time we find that there's a lot of investors in the region who do not really understand about the impact about AI. Um, not like um, in Silicon Valley or in China where they already see a huge impact. As I mentioned um, companies like Minimax and Jeepu, when they went um, public they already had a valuation of uh ah, 30 billion and above USD. Um so we have to do a lot of uh, education work. We have to convince investors and convince uh. A lot of our supporters said we have the similar kind of capabilities and we need their support because we can't be the only player who want to achieve the vision of um, building AI for inclusion, building AI for third nations. Besides the G2 of China, U.S. the third agent did own AI company. Um, so that's one of the challenges which we have trying to educate the market.

Speaker B: Yeah for sure. I mean especially when you're doing something like very much like frontier, like state of the art. Right. It's probably hard to get the investor who come along on the ride as well because I guess some investors probably they see as like oh the markets already have like you know, Z and this like frontier model already. Why do we need a company?

Speaker A: Right, yeah.

Speaker B: So it's a tricky one but I think your strategy of targeting a lot of the developing market or even restable I would say is probably pretty fair because I, yeah sometimes I remember when you said like what is it like 5 um, percent of the 10%. Right. You actually pay for the subscription like 05. Remind me. I think there was a picture online. You probably might have seen it too, right? There's like the whole world is like this map of the dot, right? The spread. I don't know if I can find the picture now but like the map of the dot and then where like the gray, gray is like, I think majority more than like 90%. They don't use any AI at all. And then like there's some green, green dot, right. People are using the free version like just like free chat. And then like the orange. Yeah, I think you know what I'm talking about. Yeah. Orange is like the one that actually user paid the $20 and then the red one is like the one very advanced user. Maybe like you and me, we use like, I don't know advanced like coding, cloud code work and a lot of the stuff. So we like just tiny little, maybe 3 million I feel like. You know. Of course obviously if you go talk to one AI startup, we leave it like our own like little AI, you know, like uh, sphere or bubble.

Speaker A: Yeah, right.

Speaker B: But then realize that we're very early, I think.

Speaker A: Yeah, totally agree with you. So I think this is a situation right now. A lot of uh, um, you know, stakeholders do not really have the um, understanding about the impact about AI. Like investors, like a lot of um, um, potential government policymakers, if they know, if they would realize um, the huge power and huge impact, huge potential of AI, they will probably spend much more money in supporting companies like us. Just like how us, if you look at us, the top 10, the companies in US right now is worth 4 trillion. 4 trillion.

Speaker B: It's huge.

Speaker A: Unbelievable. Companies like Opn, Anthropic, they are able to assemble a lot of money. It's like both us and China realize that um, this is the final say about the future development. That's why we have to squeeze some money and squeeze um, um some good amount of support for these companies. Um, just like how the Federal Reserve and China's, you know, uh, China banks um, is probably uh, you know I would say uh, printing money to support these companies, uh, AI companies so that the rest of the industry everywhere is paying about 20% of tax. Because this is how the entire, the valuation um, will shrink a little bit so that we have enough reserve to support the AI and robotics uh, companies. And I would hope that uh, Singapore and the rest of Asian countries will do the same because not only limited to the South Asia, but the rest world will do the same because we don't want to lag behind from the top two. Otherwise in terms of dominating, in terms of future conflicts, they would dominate every aspect. So uh, this kind of uh, education we have to spread to the rest of the world, especially to the stakeholders, uh, so that we get the kind of support um, to build our product really well, to spot our Vision.

Speaker B: I see. Yeah, that makes sense. Yeah. I think it's going to be a very long road obviously if you educating frontier stuff technology. Cool. So I mean obviously your company is targeting a lot of the consumers with the main one. But I'm curious, do you have any uh, future plan of maybe expanding to target business and enterprise as well?

Speaker A: So version about this April which is second quarter we're going to launch our enterprise solutions which is uh mostly covered as BaaS services which is model as a service. So we're going to introduce features like our search, our office work, our uh, uh, you know uh, image and video generation and maybe companion. All of these cover as uh um enterprise solutions. And how we're different from other models is you see that we do a lot of intended routing and we have very complex tasks, uh divided that into a smaller you know, item level unit tasks and route to different you know um, foundation models. So if companies want to tap on us to reduce their work on um building their own agent agent framework, reduce uh work of building their own SaaS services, they definitely can go after us. Another advantage is as I mentioned we have uh a huge optimization in terms of uh speed and cost. And for companies uh, especially in companies in emerging countries cost is a very strong factor because we're serving for the um, less well paid customers. So uh, um, we enter as a very strong force uh in enterprise solution starting from the second quarter.

Speaker B: Interesting. Yeah, that sounds very exciting. I know some of the other frontier company. Right. I think the enterprise solution usually boast something along the line of privacy and maybe like enterprise they don't keep the data and whatnot. Right. And then trying to blend it into. I don't know. I'm sure they probably have some agentic component in there as well too for the businesses. Yeah, it's pretty common. Cool. Yeah, Interesting. And are you thinking of targeting a very big enterprise in Singapore first obviously because that's where you base right?

Speaker A: I don't know. Uh, so as I mentioned ah, um time is the essence. If you go for big players probably will take maybe half a year, one year to close that.

Speaker B: That's true.

Speaker A: And for the time being we want to be extremely fast. We probably want to just open our APIs. Um, maybe we're not even serving for the end customer. It's more like the uh, system, system integration players, um, partners who sit on top of our system and serve for these clients.

Speaker B: I think that makes sense because at least it will help you. They already have an existing distribution, right?

Speaker A: They already have existing, yeah.

Speaker B: Exactly. So you can cut down on the sales because I know that yeah, enterprise sales cycle where there's like medium and big enterprise is like very long like cycle like what is the sales cycle. Right.

Speaker A: But you know, I mean by the time we close our first biggest player we already potential public company. So you want to focus a lot in terms of fast situation. So we get the consumer apps first and we start with set up mass services which is a model as service. We provide service to the system integrator to uh, fast builders instead of serve big clients.

Speaker B: Okay, interesting. Yeah, that's the first time I heard of that model as a service. Hopefully in the future we'll hear that more. In the future you might be the first one that came up with that name I guess. I don't know.

Speaker A: It's not a new name but it's probably a new name.

Speaker B: It's not a new name. Okay, okay, okay, okay. Yeah, I see. Interesting. And um, yeah some question about Gen AI in general. I'm curious because I'm sure you and I, we've read a lot of how the future works. So I want to hear some of your prediction here as well. One is um, well how will Gen AI evolve in the future? What is your take on that? Where do you see it?

Speaker A: Yeah, I definitely think that AI is changing drastically towards the world even right now you have seen the huge changes. But um, a lot of M Labs um um are ah holding their most advanced uh models because they can only release um the one or two generation generates down the road. Um they're not going to release their newest model. So I would say that within one or two generations later we probably will see AI surpassing AI surpassing human in almost every aspect. That's why you see that influencers like Elon Musk, Sam Altman, Demis has obviously all believe that AGI is coming very soon because for the most updated versions of the AI models probably used in somewhere else maybe in US army operations they already super advanced, super super advanced. And in terms of um, future um, of uh, um SaaS services and future of apps, I believe that most of these companies won't exist or they were consolidated into more AI centric companies is AI which dominate um the entire future. Uh anthropic cowork, cloud cowork with skus is able to disrupt a lot of industries because they don't have to exist. With AI and skus you don't have to do a lot of detailed process work because AI would handle it. We're not going to develop software for Human animal. It's mostly software for AI which is the skills and AI will handle all the um, hard and dirty work. That's why you see um, a lot of SaaS services do have the mindset that we're going to provide the most convenient product for human. Um, but the problem is human are not going to operate on the software anymore. So uh, a lot of uh, SaaS services are not going to exist. A lot of applications are not going to exist because similar um, to what I mentioned about SaaS, you don't have to uh, um there's no usage of one person using different apps to you know to serve for different purposes like um, booking a ticket, booking a hotel, um, go for a journey, uh, maybe buy some stuff from E Commerce. All this entire stuff, um, operation and interaction with the network will be handled by your agent and you just have to communicate with your agents using natural language. Just like you know how Open Cloud is already achieving part of the, part of the, part of this realization. I believe that within two or three years the world is clearly dominated by a few very big AI labs and other um, you know all the other applications and SaaS services will either exist as SKUs or um, MCP tools, uh, or ah tools connected by mcp. So we hope that we become one of these labs just in case that all the other labs are coming from China and us. Uh, we still want to be a third party on a neutral land and serving for the emerging countries um, and have the kind of uh, AI priority and we hope that more people are going to believe in our what we have seen, we have seen and believing our vision and support us and we hope that we can educate the marketing very fast. Um, um a lot of investors are still looking at SaaS services, looking at vertical AI but not only mentioned by me but also by uh, Elon Musk and Dario from Anthropic or uh, semi automated. They all know that it's not going to be the case. AI especially in general purpose AI is going to disrupt the entire, disrupt entire vertical um, vertical sales services and entire vertical uh, AI agent companies. There's no point spending uh, you know um, betting for the growth of these products even though they might have some growth within a short period of time, long term speaking is definitely going um, to be the homeland for um, AI, uh general purpose AI agents. So that's something we truly believe and we're truly spending our effort.

Speaker B: Interesting. So I think what you said to kind of remind me of this, what is it the prediction Tombsday, um article Right. Written by CC Citrini, if I remember correctly. Right. I'm sure you probably might have read that piece also the very long piece of like in 2028. Right. Which is like what you said. Two or three years down the road a lot of the SaaS, vertical companies that will cease to exist because they just cannot compete because company now kind of build their own SaaS. It's very interesting futuristic scenario. I'm not sure if it will come true or not, but I think it'd be very interesting right now.

Speaker A: It's exciting to see it's coming right now. Yeah, Just think about that in the future. It's not about making products for human, it's about making products for AI. Because AI agent is your only companion which takes order from you and handle all the hard work. So there's no reason building software, sales services, even vertical AI. Ah for human to use you just need a few very strong AI companion. Um, assistant does the job for you. Just like how Open Claw is already uh, opening the door for it.

Speaker B: I see. Yeah, it makes sense actually. One thing I've also read, um, unlike a lot of, I mean there's probably a lot of debate opinion about it, but I feel like there's something called diminishing uh, amount of return. Right. As in with the model. I mean the model is probably going to get better and better. But uh, I think what I was reading is the efficiency, it's not going to like 10x, let's say from OpenAI 2, OpenAI 3 it's like a 10x jump. Right. But then 3 to 4 is not going to be 10x, maybe like 5x or 6x. And then 3, 4 to 5 is probably going to be less than that. Right. So I mean I feel like it's going to be diminishing amount of return here. So how do you actually see within. And I know people also like say a lot about AGI. Right. But I don't think many even expert in the AI if you agree what AGI is. So I feel like. But uh, in my opinion I feel like AGI is like a huge gabber. It cannot be a 2x. It has to be, I don't know, 100x or something.

Speaker A: Right.

Speaker B: Versus a current one. So I'm kind of curious uh, how you think about that as a lot of the labs like rich plateau.

Speaker A: Yeah, there's a lot of plateau as you mentioned about diminishing returns in terms of models for sure. Because for foundation models a lot of benchmarks is already um, hitting the very Top scores especially in terms of math programming, a lot of objective matters. Um AI is already reaching um a score way higher than human in terms of some of subjective matters like uh doing a research, doing something, something with aesthetics which you have to understand. A human perceptiveness is going to be um still difficult, still have the gap. But as you mentioned the newer models are taking maybe a smaller percentage of growth because the big jumps are already done by the earlier generations. But we still see uh a lot of potential growth opportunities within next couple years. First the cost is still very high. The cost is high because of uh what we call AI economy because everything has a cost analyst is uh running um on uh Nvidia's uh GPU. While Nvidia something like 70% of the gross margin it maintains the cost very high so that it's difficult for AI inferences to uh uh do a huge amount of work. That's also part of problem with um AI parity. So in the near future I would hope that um with introduction of more um chips company um more advanced version of the uh GPUs um the cost of GPU will lower to a huge amount. Maybe one tenth of what we are seeing right now. Uh number two we see that there's still a uh problem with um uh speed of AI. You might see that token per second 150 to 200 is good enough. But if you want to really have um fast feedback, feedback loop like how human is reacting. Right. Um if you want to have the kind of uh um the feedback loops um um process brain processes like human to be injected into a ah human like humanoid um robots um you need to keep improving on the algorithms to make much stronger integration of software and hardware in terms of speed of process. And there's also a lot to go in terms of how to make this robotics um more like human by understand about the instructions and have the memory and can really get into the house code to do some work. So despite the growth of um financial model which I agree with you that return um the marginal returns is diminishing. We still see ah a huge growth opportunity in terms of integrating the foundation models into the household work. Especially with speed and cost. So that's something we try. Our team will work together with industry with uh open source community to hopefully

Speaker B: um

Speaker A: follow the journey.

Speaker B: Makes sense. Yeah, so it sounds like, I mean yeah so like from what you said that we have a huge well maybe not huge but a very small like ecosystem of suppliers for like AI equipment. For example like Nvidia GPU and you said. Right. Uh, I think there's like a M. Probably a bottleneck there somewhere if I remember.

Speaker A: Definitely a bottleneck.

Speaker B: Taiwan tsmc. Right, tsmc. Because they also supply. I forgot they supply something for Nvidia and Nvidia mainly M buy them.

Speaker A: So if something manufacture to the.

Speaker B: Okay, yeah, yeah, yeah, exactly. So yeah, I feel like. Yeah, I think they are like a monopoly in a way or maybe like majority of package. Right? Exactly, exactly. So I don't know who's come in and maybe compete with them or not. But yeah, that would be for sure. I think that would drive out like, because earlier we were saying if there's only two AI companies they will control too much power or two like dominant. What is it? AI country. Right. But then if there's like a third one and there'll be more competition open up, reduce the cost for sure. Yeah. Awesome. Interesting. And I'm curious, uh, so I read somewhere like, I think cloud anthropic. I think they're using their own model to build their own model, right? Maybe like what, like 40% or something of the code written by their own model. In your company, are you using your own model to build your next model?

Speaker A: This is possible. This is definitely possible. We haven't had the entire automation of the UM model development, but maybe um, 60 to 70% of the um model training is already uh, um automated by AI agents. Like um, if you want to do a reinforcement learning, you will need a lot of feedback from human previously and you will need that data to uh, build a reward model to help you to give a score for each of the trial and errors. But what we can do right now is instead of using human to do the label, we use agents which serve almost the same purpose of human and find a way to calibrate the process with a very small number of sample data. That way we are able to um, automate the process of uh, uh training uh, one generation above another. Um, have the entire evolution by itself because the newer version of the models would be more advanced and it served better as a judge to give scores, um, and the scores in turn will give a better um, reward for the process of reinforcement learning to generate the next version of AI models. So to make it simple it's like uh, it's like a uh, stretch right between the judge and the player. If you have a better judge, you have better player, but better judges. Together we have better players which in turn will become better judges. So this is a strategic game to have the AI to evolve by itself.

Speaker B: So basically it's like a cycle. Right. Basically the IO improve itself. Right. Learn and improve and learn. Improve. Interesting. Yeah, I think this is like, I feel like that's like the, what is it? Golden standards, right. AI chain on itself and then just improving. That's, that's really awesome. Yeah. Only you can succeed with that, Google. Yeah. Just a couple more questions before we close out the show here and then so, I mean I'm sure the building of companies is extremely hard and I think, I think on top of that, I'm not sure if you're still pursuing your PhD right now. Right. I think if you tell me correctly. Yeah, exactly. So basically. Oh, uh, man, that's a lot of, so quite, quite tough right there. Yeah. Do you have any advice for people who want to build their own company, their own startup?

Speaker A: Yeah, you know, I had a lot of experience before building this startup. I had a journey in the Valley. I had built a couple of other startups before I built this one. That's what, that saved me a lot of time, saved me a lot of uh, mistakes. Because building a startup is not easy, especially with AI as mentioned, time is essence and you have to give your, give your best. Otherwise the competition is also going to be very fierce. So my advice is going to uh, embrace the challenge, go to the market, step out of the comfort zone as early as possible, make some huge mistakes while you are young. I think I made a lot of mistakes while doing my first two startups. So for the third one, for this one I'm trying to avoid most of the mistakes despite I'm going to make a few more and learning is very important. Um, I, I, I feel that I'm, um, I have learned a lot. I have become a very fast learner, um, do a lot of feedback, um, on myself. You have to be a very adaptive learner, learner, uh, learner before you become a successful businessman because you will soon realize that a lot, a lot of things do not happen as you wish. What, what I really realized that is, is, you know, a lot of successful, a lot of, a lot of successful businessmen are also very humble people. Um, a lot of very smart people, a lot of very smart guys are not going to be successful businessmen. Um, successful entrepreneurs, especially if you do extremely well in your school, if you have very successful, uh, learning, life, career, it's very difficult for you, um, to become a very successful entrepreneur. Um, because that's something I think I heard from Elon Musk because maybe your capabilities haven't matched your ego. That's how you Cannot grow. So if you want to, the kind of advice which I would give, I think is extremely helpful to myself is keep my ego very low and become very adaptive learner. Um, make mistakes but fail fast and get back as soon as possible. So I, I believe a lot in this process.

Speaker B: Yeah, makes sense. I think you summarize it so well that I don't think I can summarize it any better. Yeah, for sure. Because I'm sure that many entrepreneur, they face a lot of mistakes and hurdles. Right. But like yeah, the faster or sooner you can learn it and fix and steer yourself correct the direction. Right. And then the faster you can grow. Yeah. So that's really awesome. Cool. Cool. And uh, yeah, final question. Where can people follow you or your company? Feel free to share any public, uh, links or handles. You can just read it all out or share the screen, whichever one you want. Yeah.

Speaker A: So you can search from App Store or App Store or Google Play with Agnes AI. Agnes a GNES is also named Agnes AI. Just search from Apple's App Store, Google Android and you find our app and try play with it.

Speaker B: Nice. Awesome. Yeah, I'll make sure to uh, put those links in the show note later on so that uh, for the show listener and also for the watchers, you can also try it out your app.

Speaker A: Thank you very much.

Speaker B: For free here. Awesome. Cool. Yeah, thank you Bruce. It's been uh, really enlightening today having you on the show and hopefully that's

Speaker A: a lot of fun

Speaker B: creating a third uh, nation, third country nation. But very will be very powerful being a computer along.

Speaker A: Ah, very good at reflective time for

Speaker B: looking forward to that.

Speaker A: Yeah, looking after it. Yeah. Awesome.

Speaker B: Thank you. Thank you so much for being on here. Cheers.

Speaker A: Bye bye.

Speaker B: And we're back to my studio. What do you think about the uh, demo that boost show regarding the capabilities of Agnes AI? Would you consider becoming a user after seeing his demo and the capabilities of Agnes AI? If you have tried Agnes AI uh, models, let us know what you think below in the comments. For the next episode we go into the intersection of AI and Web3 we talk to Gabriela Moreira, the lead developer of informal systems. She's one of the main uh, builders behind Quint, a tool that helps developer make sure complex systems actually work the way they are supposed to. Quint is especially helpful for things like blockchains, distributed system and smart contracts where small bugs can cause huge problems. If you enjoyed this episode about AI and entrepreneurship, give a like and subscribe. Thank you for watching and see you next time.