Emergent: How Six Months of Tinkering Led To A $100M ARR Company

1. 0:00 – 0:56

   Intro

   1. MJMukund Jha0:00

      If you look at like last 30 years, like most of the economic gain in the world has come from software companies. If you remove all the software companies from, uh, you know, Nasdaq and, and S&P, you'll see it's been just a flat line. And, and we started thinking, okay, what if we can bring this power to almost everybody in the world?

   2. JFJared Friedman0:15

      [upbeat music] Welcome.

   3. MJMukund Jha0:20

      Super excited to be here. What a crowd.

   4. JFJared Friedman0:21

      So Mukund, um, maybe not everybody knows what Emergent is-

   5. MJMukund Jha0:27

      Yeah

   6. JFJared Friedman0:27

      ... and also like what a big deal it is.

   7. MJMukund Jha0:29

      Right.

   8. JFJared Friedman0:29

      For those who don't know, Emergent is one of the fastest AI growing... is the fastest growing AI companies in the world, and really, I would say one of the first truly AI native companies in India to get to real scale.

   9. MJMukund Jha0:42

      Right. Right.

   10. JFJared Friedman0:43

       Um, and so you're gonna get to hear from... I really see you as like a pioneer of a next generation of startups coming out of India, and y- you're gonna get to hear how, how he's done it. Um, to start with, maybe you can just tell everybody what Emergent

2. 0:56 – 2:54

   What is Emergent?

   1. JFJared Friedman0:56

      is.

   2. MJMukund Jha0:56

      Yeah. So, uh, I mean, uh, thanks for inviting me. I'm super excited to be here. I can't, uh, imagine s- like so many people coming, uh, to the school and the whole energy in India about the whole YC trip has been amazing. I was at IT Delhi, uh, a couple days back, uh, same energy. Uh, so super excited to be here. Um, Emergent is a platform that allows anybody without any programming knowledge to be able to build software that you can actually ship, uh, that your users can use, that you can monetize. Uh, essentially we are riding on this whole wave of, uh, coding becoming easier with AI. Uh, and uh, when we started our journey, we actually started off as a research lab building coding agents. Um, became world number one on SWE-bench, which is the benchmark for all of the coding agent. Uh, it was just a four people team which actually s- got us there, and then we started thinking about like, "Hey, what would happen in the world if we can democratize coding for everybody?" Um, and me being a programmer, Madhav, who's my co-founder, who's my twin brother, um, he... both of us have been programming since age 12, um, and super, super passionate about, uh, you know, programming. And, um, one of the things that we realized that if you look at like last 30 years, like most of the economic gain in the world has come from software companies. If you remove all the software companies from, um, you know, Nasdaq and, and S&P, you'll see it's been just a flat line. And, and we started thinking, okay, what if we can bring this power to almost everybody in the world? Like there are a billion people with so many ideas, so many ideas just die because you do not have an access to sort of bring them to life. And, and that was the mission that we started with. Uh, today we have more than eight and a half million people, uh, who are using the platform. More than 10 million apps have been built. Uh, we recently crossed $100 million in annualized run rate. Uh, today one of the fastest growing startups in the world. And, and the reason is that we are able to allow people to actually really ship what they dream, and it's as easy as just chatting with your agent, and we take care of everything from hosting, deployment, maintenance of the product, um, and truly unlocking the power of, you know, bringing an idea to life with just chatting with your agent.

3. 2:54 – 3:44

   9 Months to $100M ARR

   1. JFJared Friedman2:54

      How long since you launched the current version of the product?

   2. MJMukund Jha2:56

      Yeah, we launched about nine months back.

   3. JFJared Friedman2:58

      Nine months. Okay. So keep in mind, this is basically a nine-month-old company. Tell us like about the scale that you're operating at just nine months in.

   4. MJMukund Jha3:07

      Yeah. So we have, uh, close to about eight and a half million users on the platform. Um, and, um, we are, we are well over 100 million in annualized, uh, revenue run rate. And again, like I think the latent demand in the market is, is really, really high. People... there are a lot of people who want to, uh, build software and, and so far have not been able to have the access to these tools, and platform like ours truly enables them to ship, you know, uh, uh, an idea that they have had in mind. A lot of our users are actually entrepreneurs who do not have a, a tech team and have been sort of handicapped by access to technology and, and now are able to build.

4. 3:44 – 5:00

   Why build a global company from India?

   1. JFJared Friedman3:44

      Who are your users, and also where are your users?

   2. MJMukund Jha3:47

      Yeah. So we, uh, have users all across the globe in 190 countries. Uh, when I started... when I actually... like just to give you a little bit of background, right? When I came to India in 2014, I was... before that I was in Google in, in US. And I've always had this, uh, thought that, hey, why is there no Google from India? Why is there no Facebook from India? We have so much talent, so much engineering talent. In fact, you look at the top leadership of, uh, all of these companies, you know, like, uh, Microsoft, Google, you know, uh, there are sort of Indian folks who have sort of gone there, right? And I've always wondered why is there no sort of technology-first global company from India, right? And so when I was... uh, after Dunzo, when I was thinking what, what to do next, like one of the things that, that I, I had in the back of my mind was that I truly want to build a global company from India, like just like Facebook and Google. And today we have, uh, people have been using us over 190 countries. Um, and most of the, uh, like, uh, revenue comes from US and Europe. Uh, India accounts for about 10% of our revenue. Um, and, uh, but yeah, our audience is fully global.

   3. JFJared Friedman4:45

      And I'm not sure that people know, but before you started [chuckles] Emergent, you started another company that I'm sure they all know called Dunzo, which is like a really big deal.

   4. MJMukund Jha4:56

      Yeah.

   5. JFJared Friedman4:56

      You raised, what? Like a half a billion dollars, and it was like a... it was a huge company.

5. 5:00 – 6:36

   Dunzo: The Origin Story

   1. JFJared Friedman5:00

      Yeah.

   2. MJMukund Jha5:00

      Yeah. I'm sure in Bangalore, I think a lot of people would know us. Uh, you know, uh, we were pretty popular in Bangalore. We like, um, uh, you know, at our peak we were one of the most loved consumer brands in the country. Uh, even today when people ship something, they say, "Hey, Dunzo it," and, and almost became a verb, uh, in the country. At peak we were doing about 10 million monthly orders. Uh, we were one of the first people to start the trend of quick commerce in the country, um, the 10-minute delivery trend. You know, it, it, it was, it was a pretty different journey. Like I was solving problems which were very operational in nature. Um, also like last mile logistics, you know, how do you sort of set up the dark, dark store network? Um, and, uh, the lesson that, um, you know, like, um, I, I would say was applicable there and is applicable here as well is we picked up to solve the hard problems. Uh, when we started Dunzo, there were about, um, 87 companies which were doing exactly the same thing, right? Because we had... it was very simple. You, you could just WhatsApp us and, and we would, uh, you know, we were kind of like a concierge on WhatsApp. Uh, so it was super easy to get started, but I think the hard part was the last mile. Like how do you sort of really make sure the, the end consumer actually gets the product, the product is delivered in the right state? Um, and, and we chose to sort of do that. You know, we were actually doing deliveries ourselves early on. Like I had, you know, a bike and a car, and I would just Um, in the night, get an order, I would, I would jump on a bike myself and, and go and deliver. And I think early days just doing things yourself, and, and this is one of the YC mantra, doing things that don't scale, right? Really, really helps you get close to the customer, understand the real pain point, whether there's a value or not. Um, and I think like, uh, just, just, you know, being a customer yourself or doing things for the customer really, really helps.

   3. JFJared Friedman6:35

      Can,

6. 6:36 – 10:35

   Five Startups Before This One

   1. JFJared Friedman6:36

      can we actually go back in time a bit and talk a little bit about your personal background? I learned just a couple of days ago that Emergent is actually not just your second company, but your fifth startup. This guy's actually started five startups. [laughs]

   2. MJMukund Jha6:49

      Yeah. [laughs]

   3. JFJared Friedman6:50

      Tell us like [laughs] yeah, um, maybe tell us a bit about your early career, where you grew up and went to school, coming to the US and just sort of like getting started.

   4. MJMukund Jha7:01

      Yeah. So I actually like grew up in a very, uh, I would say middle class, upper middle class family. My dad is an engineer, um, and, um, obviously, like, uh, got into engineering college, um, did my engineering. Um, always had this, uh, idea that I want to do something of my own. I actually very early on saw a lot of videos of Steve Jobs and was like really, really inspired. Um, I mean, I, I saw him, him launching the first iPhone in 2007, and that was the moment like, you know, I thought, "Oh, I want to bring something to the world," uh, you know, in, in similar fashion. And in fact, I went to, uh, Spain for an internship in 2008. Bought an iPhone. I mean, it didn't work in India, but I just bought it because I liked it so much, just bought it as a souvenir for myself. Uh, tried to hack it to make it work. Uh, and, uh, then 2009 is when I went to US to do my PhD. Um, and, um, then did an internship at Google. I liked it, uh, so much, and, uh, whatever research I was gonna do, Google had a- actually done that research already two years back, so I thought there was no point. So I dropped out of the PhD program, joined Google. Was in the search ranking team. There was a 50 people team that controlled all of Google search ranking. Uh, I was the youngest person in that team, uh, so I got a lot of liberty to sort of, you know, question a lot of things because I was, um, you know, a young person who could just, just challenge the system. And at that time, Google was very anti-machine learning. Like, they didn't want to like have machine learning in search, and I was a machine learning engineer. So, so I, I, I got a lot of, um, you know, uh, leeway in terms of asking a lot of questions, saying, "Hey, like, why are we not using machine learning here?" Eventually, like got to push some of the biggest changes in search ranking when I was there for, for a couple of years. Um, then got bitten by the, um, startup bug. Uh, left that, uh, Google. Started a company which was, uh, trying to build a group education platform where you can actually, uh, you know, bring a group, group class together. Uh, raised a bunch of money. Um, eventually, like we pivoted into a B2B software company and realized that my passion was not that. I really wanted to sort of solve education, build-- wanted to build something consumer first. Uh, so returned the money, uh, shut down that startup. Started another company into, uh, sort of habit creation. How do you sort of help people form better habits? Same time got married. My wife didn't want to move to US, so I moved back to India. And I thought I could do startup from anywhere. Uh, and I had an engineering team in New York. I was, I was working from India, but, uh, realized the hard way it's really hard to coordinate, uh, you know, without, uh, at that time. So gave that up. Um, and, um, and one of the things that sort of has stuck with me like since the beginning has been that, um, and which is sort of, you know, like over time I, I've sort of realized to, uh, you know, like do more of, is just trust my intuition more. Uh, and, um, and, um, so even with Dunzo, like I started with this personal problem that when I moved to Bangalore, like there were too many things to be done. Like, I had a car to be serviced. I had, you know, electricity, uh, to be set up, gas, all of those things. And I thought there must been easier to do this. And I just, you know, um, started a WhatsApp cha- uh, group and gave that number to a lot of my friends saying that, "Hey, if you need anything, just ping me on this group. We'll, we'll help you, uh, get that done." So started with a personal pain, pain point that, hey, like we wanted to sort of, you know, make life more convenient in urban cities. Um, and I think that has sort of stuck with me throughout, you know, that where whenever I'm-- I've been able to sort of, you know, um, solve a personal pain point, like the feedback loop is much stronger. Um, you relate with the problem more deeply. You relate with the customer more deeply. And, and even with Emergent, same thing happened. Like, you know, me and Maddy, both of us are like idea guys. Like, we have like thousands of ideas all the time. And we wanted to sort of, you know, like automate and get more of these ideas out in the life and, and, and that's why we sort of started automating, uh, programming and, and got started on the journey.

7. 10:35 – 13:21

   Lessons from scaling Dunzo

   1. JFJared Friedman10:35

      Dunzo was a huge deal. I mean, you scaled a massive company. Maybe you can remember some of the-

   2. MJMukund Jha10:40

      Yeah

   3. JFJared Friedman10:41

      ... some of the stats about how big it got. How many-- Yeah.

   4. MJMukund Jha10:43

      Yeah, so Dunzo, like we had almost a million riders on the ground. Uh, and we were doing ten million monthly orders, almost like five thousand store, stores overall. Uh, so pretty, uh, large scale, yeah.

   5. JFJared Friedman10:55

      Do you have like lessons that you took away from that experience, either things you think, you know, you did right in order to scale something so large or e- maybe even things that you would do differently a second time?

   6. MJMukund Jha11:07

      Yeah. [laughs] I mean, I think Dunzo like even though like, you know, like it has a bitter swee- sweet ending, like for us, like the, um, takeaway was, was, was like for me, were like two, two, three things. One was like solving the hard problem, right? We actually, as I said, there were like eighty-seven companies doing the same thing, and we really, really, uh, cared about the consumer a lot. Like I remember, you know, earlier-- back then there was no AI, so, so all the chatting had to be manual. And every evening there would be a spike in traffic, and every single engineer would drop what they were working on, get back on, uh, you know, on the chat screen, talk to our customers. And very early on, we had this, you know, like, um, culture where we really deeply cared about the customer. Like there was, there was a customer who wanted to ship something to a different city, and we actually put a driver, uh, the, one of the riders on a plane, uh, to send that packet. So we would go that extra mile for every single customer, and that's how sort of we, we created this genuine love from all the customers. Uh, second thing, I think like one of the things that I learned from like us not being able to sort of scale eventually was I think like focus is really important. Like, I think for us, like Dark, Dark Store was really working and working really well. But at that point, we were doing like ten other things. Like we were doing a marketplace model. We were doing pickup and drop. We were doing, you know, like bunch of those things. Um, so I think like us, like knowing that, hey, this is working, let's double down on this model would have really, really helped. Uh, but eventually, I think like I, I just see this as a series of, you know, like me being a builder, you know, um, you know, just, just as a stepping stone to, to do something bigger. Yeah.

   7. JFJared Friedman12:33

      Okay. So you worked on Dunzo for a bunch of years. You scaled it to this really huge company. It must have been a very intense experience running a, you know, a like atoms-based-

   8. MJMukund Jha12:42

      Yeah

   9. JFJared Friedman12:42

      ... business where all kinds of things go wrong every day, I'm sure. [chuckles]

   10. MJMukund Jha12:47

       Very, very, very hard. Like I mean, yeah. I mean, we had, we had a team called Watchtower, which would watch over every single order, and, um, it, it, it was like almost like a war room. You're in a war room continuously because everything... Operational things break pretty, pretty often. Yeah. And a lot of that is sort of I borrowed here, so the way we sort of run Emergent as well is we monitor all the, all the tr- uh, all the, all the tasks that are, that are getting built, all the software that is getting built, and, and if some things are breaking, we flag that. So a lot of the operational rigor I've been able to borrow from, um, you know, Dunzo to Emergent as well. Yes.

8. 13:21 – 15:44

   Leaving Dunzo and finding Emergent

   1. JFJared Friedman13:21

      So in 2023, you've been doing this for a number of years, and you left Dunzo. Um, tell us the story of like leaving Dunzo and then what, what Emergent like came out of.

   2. MJMukund Jha13:34

      Yeah. I think 2023 like, um, at one point we thought like Dunzo was too big to fail. Um, and you know, we had raised $100 million in a recent round, and I actually told my co- co-founder that, "Hey, I think now we are too big to fail." Uh, right. And, and of course, like this story didn't end, end that way. Um, so when I got out in September 23, I was actually pretty depressed. Uh, like didn't, didn't want to do anything in my life. Um, and for like first six months, I was just, you know, um, reflecting on, A, what could have we done better. Luckily, like AI, AI was happening at that time. So you know, like ChatGPT was just taking off. Um, GPT-4 had just come out. Uh, so I think it was, it was a little bit easy for us to sort of build things and, and sort of building and coding became sort of my escape from, from all the, you know, the noise that was there. So I would actually spend like 10, 12 hours just sitting on my computer tinkering with, um, all, all the things that was coming out. The new voice models were coming out. You know, people were... There were new open source models coming out at that point. So I actually got this luxury of six months of like just pure tinkering on things that I really liked with no sort of objective in mind. Um, I, I built this like, um, uh, an assistant on my Mac where it could actually talk to me and I, I could sort of... Something very similar to OpenCloud, but a very early version of that. And I, I was just following, you know, like whatever was exciting to me at that point, and, uh, it became very clear to me very early on that like coding as a space is gonna be one that, that's gonna get disrupted very quickly. Um, and I spent a bunch of time in the US with my friends, with, with people at the labs. Um, but I think it was just pure joy of tinkering, pure joy of just building something without any pressure, um, that sort of led us to sort of think of this idea and le- led us to sort of, you know, um, build Emergent, uh, in some way. Because all the insights that we got while tinkering we were able to apply while we were building the product. Um, and, and, uh, you know, that really helped. And I think just having this, um, sense of curiosity and sense of, um, you know, like when you're, you're building things just for the pure joy of it, just for the, um, you know, because, because you want to solve a problem, right? I think, I think that allows you to go really deep into the problem and bring insights that is otherwise very hard to get.

   3. JFJared Friedman15:43

      I, I like,

9. 15:44 – 17:25

   Tinkering as a Startup strategy

   1. JFJared Friedman15:44

      I kind of love this picture of you. You like, you just had this super intense experience. You build one of the top companies in India. You're burnt out. You're basically just like recuperating.

   2. MJMukund Jha15:53

      Yeah.

   3. JFJared Friedman15:53

      And in your spirit time, 'cause you have some time then, you're just like tinkering with the latest models. Just seeing, oh, maybe we could get like ChatGPT to write some code. I don't know. May- [chuckles]

   4. MJMukund Jha16:05

      Yeah. I mean, it, it was practically like just, you know, like, um, me... I, I mean, just going back to like in the old times when I was a kid, you know, like I would just pick something new and, and play with it, and it just felt like the same thing. That I was just playing with this new technology and, and, uh, the pace at which, uh, you know, models were sort of accelerating, it was, it was really, really fascinating for us to see that and for us to build a lot of deep insight into like how LLMs are gonna progress. For example, like when we started Emergent, most of the companies were building, uh, co-pilots. That that was the fashion. That was, that was what, what every VC, uh, wanted to hear. We, in fact, went and pitched to like 10, 12 VCs, got rejected from most of them. Uh, and this is, you know, Dunzo founder who was just at a big company coming out, getting rejected from most VCs because we told them, "Hey, we're gonna automate software engineering." And they thought it was crazy. Like that, you know, it was the AI is not there yet and... But we could see. We could see the model are capable. Like, you know, if you, if you just project it out a little bit, um, you know, that the steps that they are failing like could be easily trained back. Um, so we, we took this very massive view that AI progress is gonna be exponential, and we will always build in the direction of AI and, and that sort of led us to, um, sort of think from a problem perspective that, hey, let's automate all of software engineering versus piece by piece, uh, thinking of that. So I, I, I think having that, uh, downtime and just that tinkering energy like really, really helped, uh, uh, us find the way.

10. 17:25 – 18:07

    Living at the edge

    1. MJMukund Jha17:25

       Yeah.

    2. JFJared Friedman17:25

       I, I just wanna like pull on a thread from this because I think this is really good general advice for everyone in the room. Like what, what Mukund was doing, we, we, we have a name for this at Y Combinator. We call it living at the edge. It's like the models weren't good at writing code yet.

    3. MJMukund Jha17:39

       Yeah.

    4. JFJared Friedman17:40

       And when you like pitched to VCs, they were like, "The models like aren't gonna be able to do this."

    5. MJMukund Jha17:43

       Right.

    6. JFJared Friedman17:43

       And like they weren't quite able to do it yet, but you could tell that they were... That like if you, if you projected out it was going to work.

    7. MJMukund Jha17:49

       Yeah. Well, you could see the sparks. Yes. Yeah.

    8. JFJared Friedman17:50

       Right. And like that's just where a lot of the best-

    9. MJMukund Jha17:52

       Yeah

    10. JFJared Friedman17:52

        ... startup ideas come from. It's the things-

    11. MJMukund Jha17:54

        Right

    12. JFJared Friedman17:54

        ... that aren't quite possible yet. That's maybe a good segue to talk about some of the technical details of Emergent. Like, um, if you just go to Emergent, maybe you don't realize the sort of like deep technical foundations that it's built on. Can you talk about that?

11. 18:07 – 20:42

    The multi-agent architecture

    1. MJMukund Jha18:07

       Yeah. So I mean, we actually, uh, you know, like when we started our journey, like most people were building co-pilots. We thought we'll build autonomous agents that could do... A- agents was not even a word then. Now it's obviously everywhere. But like we built this multi-agent orchestrated system where you have, uh, different agents who will... Which will come in different point of time, uh, and, and perform different- Um, actual like for example, we have an automated testing agent which will test your app. We have a design agent that'll design your app. Um, all of this is coordinated, you know, through a, a large memory system that we have built, which was sort of self-learned. Every time a new app gets built on Emergent, like, you know, our agents actually extract from that what are the learnable aspects and sort of store it in memory. So every, every new app actually getting built on the Emergent makes the platform even better. Um, and a lot of the energy has gone into us into collecting a lot of the data that we have now. We do a lot of RL on top of that. Uh, we do some amount of fine-tuning and, and but a lot of the things that we have built essentially is all of the infrastructure that we have built, uh, ourselves. So we have built all of the coding agent. We have built, um, all of the infrastructure. For example, we-- when we started, there was nobody building, um, deep container technology, so we had to invent a lot of the container technology ourselves. Like for example, uh, we wanted to preserve state so that you could have multiple parallel agents running on the same, same snapshot. So we had to invent disk snapshotting, memory snapshotting, all of those things. Uh, uh, and I think one of the things that you, as you said, like, you know, living on the edge, you actually discover these problems much early on before, you know, like other, other, other ecosystem discovers it. And oftentimes you'll have to go solve it thems- yourself. Like for example, today, like we have, um, multiple different sort of parallel agents that can sort of swarm together and, and complete a task, which we think is gonna be like, like, like the future. And, and what we are observing is that every time a new model comes here, comes out, like for example, a new class of model, for example, Opus is a new class of model. Like you have to actually delete whatever you have learned so far and sort of reimagine the world from the lens of this new model. Uh, so, so far, like, you know, in nine months, we have already rewritten our system three times. Um, and, and just, just when a, when a new model comes out, we have to sort of start rethinking that, okay, what are the new possibility that's gonna open up and what, what, where this model is gonna be in six months. Um, and, um, and one of, one of the things I was telling you before that, you know, like that, like when we started Emergent, like one of the biggest challenge was actually that models could not do a good JSON output. Uh, and like there were like at least twenty or thirty YC companies that were solving the exact same problem, JSON parsing, right? And we took this view that, hey, like, you know, like the next model will be able to solve this. So let's like we just completely skip that problem, start building the agent and, and sort of, you know, went on the journey. So I think living on the edge and just trying to, uh, imagine what is possible in the next six months is really important as you sort of progress through your startup journey.

    2. JFJared Friedman20:41

       Um,

12. 20:42 – 22:42

    Beating the SWE-bench benchmark

    1. JFJared Friedman20:42

       can you talk about beating the benchmark? 'Cause that's a, that's like a-

    2. MJMukund Jha20:45

       Yeah.

    3. JFJared Friedman20:45

       Like a core, core part-

    4. MJMukund Jha20:47

       Right.

    5. JFJared Friedman20:47

       Core part of the founding story here.

    6. MJMukund Jha20:48

       Yeah. So, um, so one, one of the things that like happened when we went to YC was that, uh, and this happens with a lot of YC founders that, that, you know, like you, you come in with a different idea. You sort of, you know, stumble upon a different idea. When we actually went to YC, we were building testing agents initially, right? And, um, and, and, and when, when sort of we were coming, coming from India, like we drew, drew this on a whiteboard that, hey, like very soon you'll be able to build, build web apps, mobile apps, um, you know, through AI. And we had this diagram that, hey, like we'll be able to build web app, mobile apps on, on, on this thing. We-- day one, we went to our YC partner, told him that, "Hey, we want to build a consumer app building company." And they said, "Okay, this, this, you know, like maybe you should think about enterprise. This, this seems too ambitious." Um, and for the first, like it's a three-month program. So for, for, for three months, every week, we would have a new idea on the board. Okay, idea of the week is, you know, let's say AI Zapier, and, and we'll spend a week sort of, you know, building that or, or tinkering with that. Um, and eventually, like, you know, every, every week we'll have a new idea. We were pivoting like crazy and, um, and team was getting frustrated. "Hey, like, you have a new idea every week. What are we gonna do?" Uh, so almost just to distract them, I actually picked this benchmark SWE-bench, which was the hardest benchmark at that time. And I told them, "Hey, like while I figure out what, what we're gonna build, let's just attack this benchmark because, you know, it'll allow us to solve harder problems." And, and so almost send them in that direction. It took us three months to sort of crack that benchmark, became world number one on that benchmark. But that's really set us the foundation for Emergent, where we were able to build world's best coding agent. Uh, all of the innovation that we sort of have in Emergent right now, whether it's, it's parallelized test time compute, um, all of the memory agent to agent communication, a lot of those things we were able to discover when, when we were on this benchmark. And I think like, like even today, I think like, um, at-attaching yourself to a number which, which can sort of show you progress is really, really good way to sort of, you know, um, attack a goal or, or go, go towards, uh, building a company because that sort of focuses you into right direction. It gives you like a really good feedback in terms of what's happening. Yeah.

13. 22:42 – 25:32

    Second mover advantage

    1. JFJared Friedman22:42

       Yeah. It's super, super impressive what you guys did beating, beating that benchmark before you even really had a startup idea-

    2. MJMukund Jha22:49

       Yeah.

    3. JFJared Friedman22:49

       Like for what to do around it. Um, a, a recurring theme of the talks today has been, um, this concept of second mover advantage. Um, you know, like Zepto wasn't the first grocery-

    4. MJMukund Jha23:01

       Right.

    5. JFJared Friedman23:02

       Delivery company, and Giga wasn't the first AI customer support thing. Emergent was also not-

    6. MJMukund Jha23:07

       Right.

    7. JFJared Friedman23:07

       The first AI website builder-

    8. MJMukund Jha23:08

       Yeah.

    9. JFJared Friedman23:08

       To launch. When you launched Emergent, there were already a couple of like pretty big players and probably a whole bunch of-

    10. MJMukund Jha23:13

        Right.

    11. JFJared Friedman23:13

        Small ones. Um, much like, I guess, your story of starting Dunzo when there were already eighty-

    12. MJMukund Jha23:20

        Eighty seven.

    13. JFJared Friedman23:20

        Eighty similar companies. What, what, what gave you the confidence to launch this anyway, even though you weren't the first to the market, and how have you been able to carve out such a, such a big space for yourself?

    14. MJMukund Jha23:34

        Yeah, I mean, for us, um, when we looked at, uh, the problem space, we realized that like most of the other platforms that were out there, like they were mostly focused on front end and building demo ware, right? And, and that's where like where a lot of these were finding product market fit, right? But what we realized was that like users are actually gonna want real software to be shipped. Um, and, you know, the problem is far from solved, right? Like, and we saw that the, the expectation that user have versus, you know, and I'm sure the same thing is with Giga as well, right? Like the expectation that user has like, hey, my queries get solved. Same expectation our users have that my software should actually work right when I, when I'm prompting something. And most of the solutions out there, like even though they, they were like good at getting started- Right. They're like really bad at finishing. Like, you, you will not get a working software out of that. You will not get a... You will not have a real back end. You will not have a real databases attached. And, and so we came to this from, from a very different angle, saying that, "Hey, if you were to automate all of software engineering, how would you approach the problem?" We almost built everything ground up. Um, and we could see, like in practice, when we sort of ran prompts on all, all the platform and asked, like we were massively outperforming everybody else in the market, right? So that allowed us to sort of really attack, attack the market in a big way. But I think again, like we, we came to this from a very sort of consumer insight that consumers are actually gonna want real software that is working and not just prototyping demos. And, and nobody in the market was solving that. Um, and there were no good solutions that actually could, could take you to the finish line, and that's why sort of we, we attacked that. And once we had the product, like we had to think through GTM. How do we market it? Uh, we looked at like which companies are growing really fast, what, what, what they have done, um, and, uh, sort of almost converted our growth into a maths problem, saying that, "Hey, how many social views do we need? How many like, you know, um, impression do we need? How many clicks will we get? How many users will we get?" And at that point we knew, okay, like influencer is a good strategy for us to sort of really launch because we knew the product is really good, working really well. We just need to get it in front of as many users as possible, and that's sort of been the growth engine for us.

14. 25:32 – 27:11

    Building global from Bangalore

    1. JFJared Friedman25:32

       Where is the Emergent team based, and how do you think about building an AI-native company that targets a global audience, um, here?

    2. MJMukund Jha25:41

       Yeah. So most of the team is actually in Bangalore. Uh, we have ninety-five percent of our team in Bangalore, uh, pretty much built out of India completely. We have a very small team in SF. We have recently opened a new office in SF, so small team is there. And by the way, we are hiring, so if people want to, you know, uh, apply and, and work at a strong AI native company, please write to me, mukund.emergent.sh. Uh, happy to take a look at that. And I think like one of the things that I've, I've realized, uh, you know, like and we generally hi- like hire for like learning slope, people who are like really passionate about, you know, solving a problem, people who get excited about, uh, you know, solving some of these problems. And what we have seen is, is that I think like one of the things that separates us right now from, from other companies is that everybody in the company generally enjoys solving and working with AI, right? I think there's this added... Of course the growth is great and, and, you know, we, we get to solve real user problems, but I think just the, um, you know, the, the complexities of the problem and the possibilities are so much that we generally enjoy like day-to-day problem-solving with AI right now. So that's amazing.

    3. JFJared Friedman26:38

       You've had a chance to build two very different companies. You built Dunzo in the sort of like first wave of great Indian startups that were building things like Zepto, a lot of like local, well, local stuff. Um, and now you're building Emergent, which is like the part of this like second wave of AI native companies post-ChatGPT. I'm curious, first, what are, what are some of your, your takeaways from building those two kinds of companies? And then second, what would your advice be to folks in the audience who are thinking about where to look for startup ideas and what kind of things to build?

15. 27:11 – 29:01

    Outro

    1. MJMukund Jha27:11

       Yeah, I mean, I think, I mean, my realization after building the two companies is that like building a company for, for India, a local company versus building a global company is actually exactly same effort. You know, it's, it's equally hard to build a company in India versus building a global company. And so my advice to a lot of people right now is just think global from day one, because I mean, it's gonna be equally hard to build both the startups. I mean, and, and it's kind of like a, like a prevalent wisdom that actually starting a harder idea is easier because you can inspire a lot more people to go after a harder problem, right? And you can sort of, you know, uh, inspire yourself to, to go after this. So, so I would, I would, I would recommend that like think global from day one because like now you have the reach, the access. Internet is with everyone. Technology is a big leveling, um, you know, for, for everyone. Everybody has the same access to same technology. And, um, and you can actually just re- reach global customer from day zero from, from India today. Um, the other thing I would say is that, one, I think, um, just following your intuition is, is really, really, uh... I mean you'll get a lot of advice, but I think following, as a founder, following your intuition, uh, is actually much better because, you know, like you probably have a better sense of general, you know, what your customer wants, um, what your customer needs. Um, and also I think, um, just thinking big and ambitious. I think whatever you are thinking right now, just ten times that, hundred times that, because I think the next, uh... You know, with AI, I think a lot of, lot of things are changing and, and it's, it's not a time to sort of, um, attack the floor. It's the time to attack the ceiling and think really big. And the bigger you think that the, the... I would say the higher probability that you'll get to success.

    2. JFJared Friedman28:44

       That's an amazing piece of advice for us to end on. Mukund-

    3. MJMukund Jha28:47

       Thank you.

    4. JFJared Friedman28:47

       You're an inspiration to us. Thank you so much.

    5. MJMukund Jha28:49

       Thank you. Thank you so much for having me here. And, and the energy is electric here, and I'm looking forward to lots more of Gigas and Emergent coming out of India over the next year or so and looking forward to these people here. Cheers.

Episode duration: 29:04