Full Transcript

Utsav Somani: Hello, listeners. It's the 1st of April. We're not going to be doing anything foolish on the show because we've got two serious people who are going to come and break down different industries. We've got Rishad who's going to talk about his fund and he's been investing in deep tech and AI. And we're going to talk about Kind and Merlin. I mean, two founders building in this space, which is pretty exciting and I think very, very topical as well. So let's get down to business. Let's welcome our first guest, Rishad. Welcome to the show.

Rishaad Currimjee - GP, Golden Sparrow: What's up? How are you doing?

Utsav Somani: Good to have you. Dialing in straight from Mauritius? Absolutely. It's nice and sunny here. So let's start off with the backstory. One 40-year-old business, family business, you left that, made a decision to come to India and launch a deep tech fund, investing in AI and deep tech way before everyone was talking about deep tech like they are today. What went through your mind to come to this decision and what was the conviction building journey like?

Rishaad Currimjee - GP, Golden Sparrow: You know, when I think about what my great grandfather would have done when he left a small town in Gujarat 140 years ago to journey to Mauritius, I sense some of that in what I'm trying to do by coming back to India and finding phenomenal talent who are building in deep tech. But this story didn't start three or four years ago with Golden Sparrow. I took a decision back in 2004, in fact, to leave New York and move to Delhi. I had a chance meeting with Sunil Mittal and he was building out Airtel at the time. It was much smaller, but he invited me to come and join him and work in M&A. So I spent a decade in Delhi. Your family seeded the business also? We had seeded the business, but we exited most of our stake a few years later. But you don't need to remind us of that. So we had a good relationship and therefore that helped me enter India, be part of this great growth story. But I've always wanted to build something by myself. So I left Airtel and did a couple of startups. One was a failure in infrastructure. The other one was in biotech. I had a small exit. And then I started investing itself. Back in 2012, I made a dozen investments in India. At the same time, I was working in the family business in Mauritius. I grew back there. But nights and weekends with Zoom happening and the whole advent of solo GPs being able to do so much more thanks to the list, you allowed me to do what I do. I decided to launch my own fund and that happened four years ago. And that's how I got here.

Utsav Somani: So 2023, you start off with a fund one, which is 8 million. And what was the thesis then? And how has it changed now?

Rishaad Currimjee - GP, Golden Sparrow: No, excellent question. So originally we wanted to invest in DTech, AI and SaaS. I say AI with... While Chachipiti wasn't what it was, I had been investing in the earlier versions, not of transformers, but in fact of the predecessors of that in some fashion tech AI companies, GAN networks. So those got me excited about AI. And I said, well, that's what I was going to do. Very quickly, we made a few SaaS investments and started to move away from that. We realized that that was not the way to go. So we went from roughly 30% DTech to 50% DTech. We did the balance in enterprise AI companies. We invested in 18 companies, $350,000 average check sizes, average ownership of 2.5% across semiconductors, robotics. We invested in biotech, biomanufacturing, quantum computing. In fact, we did half a dozen investments in DTech. I think before DTech was even the flavor of the month.

Utsav Somani: I honestly just remember one fund and the founder of that fund also came on the show, Speciale Invest, Vishesh. And they've done exceptionally well in terms of deep tech investing, but they've been sort of like this lone warrior in this whole industry. And you've backed a company called EtherealX as well. The founder was on a show to talk about his company.

Rishaad Currimjee - GP, Golden Sparrow: You know, I would say while Speciale is a lone wolf and they've clearly laid the ground for us, they were not alone. There were a few other funds out there. I would say there would be- I think it was Pi Ventures.

Utsav Somani: I think it was one of them. Pi Ventures. Yeah.

Rishaad Currimjee - GP, Golden Sparrow: I would say even Yournest was perhaps less well-known, but equally successful in terms of their presence. In fact, half of the deals that we've done have been with Yournest, Pi, and Speciale. I'll share a story that, you know, talk about Speciale. We invested in Morphing Machines, which is a semiconductor startup coming out of IISE. Two professors who said, well, let's make a reconfigurable chipset as opposed to application-specific. And that was such a hard problem to solve. The timing was now. But when Speciale came in, they said, well, I don't- we think maybe you should have a CEO. So they brought in a CEO from Stanford with a master's degree. They cleared the cap table, gave him 25% of the company, and we could invest it with them. The company has now raised a Series A. We're excited about where that's going, but I must say, I owe a lot to those few funds, but there are still very few deep tech funds. And that's the opportunity that I saw when I started investing in deep tech.

Utsav Somani: I mean, you made, I mean, approximately 18 investments, I think, from the fund one. And India VC is probably around 15 billion to maybe 20 billion. I think the number keeps varying between that. When you started, I think 1% of- less than 1% of Indian VC was going into deep tech funds. Now it's probably closer to like 2 billion, so 15% of the total overall world of Indian venture capital. And you've said publicly that founders should ask who actually deployed capital versus who is riding the wave. That's a fairly pointed statement. Does crowding of the space change how you think about deep tech investing, or you still play your own game?

Rishaad Currimjee - GP, Golden Sparrow: It's a very important question, and to address that, one should also understand what is happening and why people are crowding around. There is what you call free money coming from, you know, $10 billion, and everybody wants to qualify and get access to that. But I think government's going to be smart about it and give those who have experience and who put money where their mouth is, like the Yonais, especially as the Pais and others, who've actually put time and effort into building a deep tech practice, and perhaps some of the larger funds who've been around for so long who have experience across fintech and other spaces but have dipped their toes in deep tech. Well, we don't qualify as Golden Sparrow as a U.S. fund for the RDI money, but I think I'm hopeful that money will end up in the right places and the right flow. That means it won't come too fast. And when founders will ask themselves who should I take money from, I hope they'll turn around the leaf and say, well, who's actually invested in deep tech? Do they have the experience to understand what I do and support me in my journey?

Utsav Somani: That's fairly interesting. And India's also made some relaxations, right? For deep tech startups, there's increased the age of the company. Startup recognition period has gone up to 20 years. There's a revenue threshold that's tripled. And there's also the Deep Tech Alliance that came in. There's the India Semiconductor Mission 2.0. And then, of course, the India AI Mission as well. Do you think the policy finally matches the opportunity or there is still some gap? I used the example of the motion machine. And I want to hear your unfiltered version. What you're seeing on ground, I think this will be helpful to founders tuning in and just generally for people to be more aware, like where's the gap between policy to execution?

Rishaad Currimjee - GP, Golden Sparrow: You know, obviously, when I came in three years ago, I was skeptical. And, you know, it was all the US fund investing in Indian companies. I said, you know, I'm not going to count on anything coming through. But think about both the machines. Example I used early on, they're getting grants from the government as part of these programs. That, you know, that 10,000 crores has been earmarked for direct investment in companies. And we're seeing some of that trickle in from the National Semiconductor Mission and otherwise. So it's real capital.

Utsav Somani: Okay. And Semiconductor Mission, I've actually asked Chad Kibir to make this 90-day learning plan. I'm on almost day 45 of this. So trying to learn personally the India Semiconductor stack. I mean, the global stack, but also how it applies to the India story as well. And we've had some founders on the show, OptoML, and a few others who came on the show and educated us about the semiconductor space as well. But where do you think an opportunity for a startup lies? Like you're a solo GP fund. The, I mean, the visibility towards an exit or even a reliable outcome is probably at least a decade out, if not more. And there's a huge amount of capital required. So where does the opportunity lie for founders in the semiconductor space and what are spaces that you're seeing which are exciting to you? From packaging to testing to OSATs to fabrications to like, just let's, I mean, spend a five-minute time on the semiconductor space with you.

Rishaad Currimjee - GP, Golden Sparrow: You know, I'll share a story. In January, my soon-to-be principal, Ritik, was about to join us on the 1st of Feb. And as part of our getting to know each other and we'd worked together in the past, we were sharing various deals and he bought a semiconductor deal to us, which is a software company that helps who wants to displace cadence and synopsis. So that's more on the design and the testing side, which I think is one of the key and critical areas where semiconductor can do well in India. It's the software areas. And we decided not to invest. We didn't have the conviction levels that we needed, but he was banging the table very hard. And I said, listen, you're not even part of the team and we're not going to do this. So what I did, I flew to Bangalore and I said, let's rework how we make decisions. So I ended up giving him a Joker card. Every year he has one Joker. He can make an investment no matter what.

Utsav Somani: I read your LinkedIn post about this before the show.

Rishaad Currimjee - GP, Golden Sparrow: And this is something I think says a lot about how we're managing people. But back to semiconductors, I think software is going to be the way to go. There are obviously some great semiconductor companies that are more on the design side, fabulous, that we've invested in as well that are going to do well. But I think a lot of the talent has been that a lot of the companies we're seeing emerge on the software side.

Utsav Somani: The software side. So let's talk about the software side and the India AI mission as well. Or were you completing your thought?

Rishaad Currimjee - GP, Golden Sparrow: The last point I'd make is that, OK, you see a lot of money going into manufacturing of semiconductors in India. But let's be clear that those semiconductors that are going to come out are going to be 30 nanometers in size. They will have different functions. They have different uses. India needs to have indigenous supplies. But the cutting edge two, three, four nanometer designs and fabrication will not be done here.

Utsav Somani: Talking about software and AI, India is promising close to 60,000 GPUs as well of compute in the India AI mission as well. So where do you think the India story lies in the world of AI? Is it applications? Is it foundational models? Someone's doing work there? Or is it just building more sort of India for India solutions where you're seeing these voice AI models which are for different vertical solutions? We have the founder of Bolna on the show as well who's talking about BFSI voice agents and has scaled beautifully as well. So where do you think the India AI story is headed?

Rishaad Currimjee - GP, Golden Sparrow: I think it's a great question. We're still figuring it out. I'll tell you this, when we started a few years ago, we said we'd only invest in Indian founders and based in India. You asked me what had changed and we've started to move a little bit away from being so focused on backing India for India talent because look at last year, we had $200 billion of spend on enterprise AI solutions coming from US companies like McDonald's and Coca-Cola's of the world. India had $2 billion. So if you're going to build AI software and if you're going to sell it, you have to be in the US. So I think on the enterprise software side, I think it's a necessity to be in the US and we encourage our founders to start off there or move there or have an office there. On the foundational model side, I was talking to a founder who was building a foundational model just a few days ago. It's close to Saurav Ram. They've got a lot of GPUs coming in. The military, the DoD, a large number of companies that are part of India will not use, not allowed to use any of the foreign NLMs. Aspect of it, yeah. And it has an impact. The second aspect of it is exactly what you were saying is that there's 50 plus languages that could actually be used for voice AI in India. And whether it's leveraging Basha or building and training your models on Indic languages is going to be something that is going to add stupendous value to at least our local businesses. And I think Saurav Ram is leading the way there.

Utsav Somani: So talking about AI and physical infrastructure, you've made an investment there as well, right? I think Myolab and GPU virtualization with hosted AI. So you've made investments at the intersection of physical and software with AI. Talk to us a little bit about how you think about that space.

Rishaad Currimjee - GP, Golden Sparrow: Myolab is a IIT KGP gang that's building out of New York. So think about everything that's being built today with AI. It's using language. It's a large language model. But tomorrow, everything that is outside of what you see, what you hear will be using physical models. And so Myolab focuses on movement in time and space. And that can have applications for sports tech. It can have application for preventive healthcare. And that foundation model would not go for consumer applications. There'll be a B2B play to sell their model for others who want to build on top of it. Hosted AI is just a fascinating concept. Think about, you know, imagine your tier one data center today in India and you're sitting on a thousand GPUs. You know, one's going to Utsav, the other one's going to Rahul. And then, well, you're not using it all the time. You're using during the day and Rahul's not using it at full capacity because he has a constant use of it. What if we virtualize it? What if we created a hypervisor around it and we spice it up? Just like what happened for CPUs, we can do that for GPUs. And this company actually raised $21 million from Creandum, which is a large European fund, a few weeks ago. And so we're very excited about what they're building. The entire office is in Bangalore, but the team's global.

Utsav Somani: Nice. And let's talk a little bit about your fund investing journey. Fund one, 18 investments. How's that portfolio performing and how are you transitioning to fund two?

Rishaad Currimjee - GP, Golden Sparrow: So 18 investments. We've got three winners coming out, which is Ethereal X, Morphing Machines, Posted AI. There's around a dozen, which are still figuring it out, between half a million and a million dollars, or on the deep tech stages, more at like TRL six or seven, going to the next stages. So we're positioning ourselves well. I would say we're in the top 15 percentile of performance for fund size and vintage, which leads us to a 20 million fund too, where we want to make 22 investments, slightly larger, but still remain very much pre-seed AI and deep tech company.

Utsav Somani: And how do you think about exit timelines? I mean, I know it's a very forward-looking statement or a forward-looking question, but super long maturity cycles for these companies, right?

Rishaad Currimjee - GP, Golden Sparrow: I'm fortunate to have LPs who have that same mindset, even though there's 62 LPs in fund one and fund two will probably have even more. That's something that I really put on the table very clearly from day one, that this is a 10-year chain. And obviously we'll have exits before that. And Ethereal X, I think, is going to be a 10 or $20 billion business. And even earning half a percent of that is going to return five to seven times the fund. So we're looking for these outliers. We're looking to really hit it out of the park with each of our investments. I think this is what VC was supposed to be, to take frontier challenges and bring them to society. And that's what gets me very excited itself.

Utsav Somani: Yeah, I think we've gotten too busy with quick commerce and everything, house help and delivering food with 10 minutes and all of that stuff. But, I mean, SolarGP, 20 million, do you think the model still works? I know you're bringing in venture partners as well to add in some special expertise on your bench as well. So how are you expanding or thinking about growing the partnership line?

Rishaad Currimjee - GP, Golden Sparrow: That's an excellent question. And I would say SolarGP is what I started off doing. But today I'm a single GP in a firm which has a principal who's on a partner track. He's the fresh age of 25, but that's the type of bet we take on our own founders. Why not take it on our own people? We've got two operating partners, one in India, one in the US, one for Deep Tech, one for AI, who play roles in supporting the businesses. Then we've got a couple of experts. Anand, for instance, who's a professor of AI at Carnegie Mellon, et cetera. So we have this bench strength and the way in small funds where you have very little capital from fees, I've given up more than 40% of the carry of the fund with this team. And so we're giving the best shot possible at doing what is fair and justified for founders that we invest in to have the best possible bang for their buck with us.

Utsav Somani: And I want to get some nuggets of wisdom out of you. You said India is no longer a developing economy, it's a developer economy. You've also said that next 20 years you'll see a lot of talent stay back and build for the world from India. If we were to fast forward to 2013, what would Golden Sparrow look like and what would be the state of India's deep tech ecosystem then?

Rishaad Currimjee - GP, Golden Sparrow: No, I think it's an excellent question. I've said this and I'll say it again, is that I think real value creation for LPs and for venture firms lies in remaining relatively small. And I think maybe 40 or 50 million is an ideal size, but I think anything beyond that makes it much harder to return funds. I hope by the 2030, we'll have more than a dozen deep tech companies that have hit more than a billion dollars in valuation. Today we have few, if any. And I think that's something that I'd love to see happen.

Utsav Somani: And what's the ideal founder profile? I know that's very hard to paint, but somebody who should take money from Golden Sparrow, how does that person look like? Or maybe we can do it in retrospect where you look at the portfolio and what's been the ideal archetype of founder that has performed well for Golden Sparrow?

Rishaad Currimjee - GP, Golden Sparrow: You know, Utsav, you had somebody on your show a couple of weeks ago. I think it was Akshay from Hummingbird. And I'll say this with...

Utsav Somani: One of the deepest thinkers in the VC space. Do you know him?

Rishaad Currimjee - GP, Golden Sparrow: Yes, I do. And we've exchanged quite a bit. In fact, when we start talking, we talk about everything but the VC ecosystem. And so I'll say this. When I started off, I must have read everything they published perhaps 10 times over. And with them and with others, as much as I come from the financial industry, I actually majored in psychology. The questions that we ask our founders are quite different from I think what one would expect. We talk about what they're afraid of. We talk about what their childhood was like.

Utsav Somani: Yeah. Hummingbird actually asks that question a lot. Yeah.

Rishaad Currimjee - GP, Golden Sparrow: What drives them to build? I'm not looking for trauma. I'm not looking for, you know, the next extreme, somebody on the extreme like Elon Musk. But there is a certain amount of resilience that we look for. And we have to unpeel the layers to get there. Clarity of thought, ability to convince, all of that is part for the course.

Utsav Somani: Do you think they have a chip on the shoulder as well? Like some of the best founders? Do you see that to be true?

Rishaad Currimjee - GP, Golden Sparrow: I think chip on the shoulder is one part of it. Spike is another that we look for absolutely. And we actually have a scoring sheet which has ability to execute, clarity of thought, the ability to build teams, the spike, you know, what is the self-awareness? Are they able to convince people? What is the co-founder of relationship like? What's their empathy? And resilience. That has the strongest weightage. Resilience is how well do they know themselves and why are they building what they're building and going and unpeeling the layers to get to that for us is the biggest signal of why we would invest. And frankly, we'll do our due diligence on, you know, on moving machines or Ethereal X. But we can't, you know, we can't do re-entry rocket math. We're measuring resilience at the early stage.

Utsav Somani: Do you use Jack GPD or Gemini or Claude to help you in diligence? Like, is that becoming part of your workflow for like these deep tech diligence of these deals and opportunities?

Rishaad Currimjee - GP, Golden Sparrow: It is. We have, I think our own mind has the limited ability to capture every detail. And what we've done as of the last couple of months, we fed our framework into Claude. And we say, okay, here's how we evaluate founders. Go back and look at the last 50 founders we've looked at and tell us how you would have rated them. And we compared to which ones we invested in. It was a perfect match. The top 10 for Claude was the top 10 for us. So from there, we got confidence to say, okay, let's have Claude as a third person in the room, but let's make sure that we come up with our own conclusion first and then bring in Claude to say, well, how would you have rated this? And then we reassess and see whether we make any changes or not. And it's getting to be more and more precise because the contextual memory that you have around Claude is improving phenomenally. I think that's what they're good for.

Utsav Somani: Awesome, Rishabh. I think this has been phenomenal. Now I'm spending the next 20 minutes with these two founders who are slightly ahead of the curve as well. They built Merlin, which is a Chrome extension, and they scaled beautifully. And this was a Chrome extension that got to 1 million plus users in less than a year. And CharityBD wasn't popular. I mean, the interface that we know of CharityBD now wasn't there. So these guys scaled beautifully, and now they're doing this company called Thyme, where it's, I mean, a scary world, where it listens to everything that's happening around you on your iPhone, and it's the co-founder for your life. That's their tagline. It's a beautiful website as well. I got so intrigued by it that I wanted them to come on this show. So yeah, that I think sets up the next segment. And thank you again. Have a wonderful time in Mauritius, and hope to see you back in India soon. Thank you for the time today.

Rishaad Currimjee - GP, Golden Sparrow: Thank you, Rishabh. Take care.

Utsav Somani: Bye. Bye-bye. All the best. All right, listeners, I think I've already done the intro for the next segment. We've got Pratyush and Siddhartha from Thyme and Merlin. Guys, welcome to the show.

Pratyush Rai: co-founder - Merlin AI, Thine: So this show is one of the most exciting things right now in the Indian ecosystem. Very excited to join.

Utsav Somani: Is Thyme listening on the show? Will it give us a summary after this? It will. So I'm just turning it off. Let's do it. Let's do it. So what did I get wrong in the intro? Did I describe it correctly, or would you want to redo it over the next minute?

Pratyush Rai: co-founder - Merlin AI, Thine: I think you got it almost perfectly right. We started our journey with Merlin, trying to build an early co-pilot for browser, then trying to become an all-in-one assistant and be scaled to 20 million users at RP. Even today, more than a million weekly users on Chrome have Merlin as one of their primary AI extensions. As part of our journey of building Merlin, what we realized is that the most important part of anyone's journey with AI is how well can AI know them. One of the things which Rishabh also talked about that he wants Claude to be in the room in some form, and that exactly was the insight with which we started with Thyme. Today, you have an AI which is among the smartest developers, among the smartest thinkers and so on, but what is the biggest problem with any AI you use today? It's only limited to the prompts you're sending it to. It does not know your world beyond the prompts. So we thought about how can we let AI have that deep of a context as your co-founder would or as your co-host would, and that's how we got started with Thyme.

Utsav Somani: Should we start with Merlin first? I think I'm very curious because I'm always fascinated by founders who are ahead of the curve, and you launched this in 2021 when CharityBD was even not out there, and CharityBD went viral I think late 22, 23, and Merlin was already in the market, so that timing was ahead of the curve. How did you think about it? What was the motivation? What was the thinking process? How do you discover these white spaces? Because I think most of financial returns in life or at least success in life lies in being ahead of the curve, and you take enough of these shots and you're already Elon Musk, but how do you think about these things? How did you come to them?

Pratyush Rai: co-founder - Merlin AI, Thine: Yeah. So one thing that worked in our favor was that we were already extremely deep into the Chrome extension ecosystem. Before Merlin, we had launched a note-taking Chrome extension where you can like, you know, along with your Google Meet, you can write notes and so on, and we were trying to build integrations between Jira, GitHub, and so on, like other developer tools using Chrome extensions. Just one correction, we launched Merlin in 22 itself. Slightly, you know, in the early days of CharityBD when just CharityBD was coming out. So we were then... This was 3.5 then? Or what was the number then?

Siddhartha Saxena: co-founder - Merlin AI, Thine: 3? 3.5. So it was 3.5 turbo at the time. I even remember like, you were trying to reverse engineer a lot of things from like CharityBD's API calls that were going through and like trying a bunch of those things. But yeah, Pratik, show us again.

Pratyush Rai: co-founder - Merlin AI, Thine: Yeah. So in the early days, basically what gave us that foresight was trying to live in the future, right? Like we always thought that people are going to, you know, give context to AI in a lot of different forms. Originally, when we were working with developers, we thought that, okay, developers are going to give context around their GitHub repos, around their Jira tickets and so on. And then AI is going to, you know, automatically fill on their Jira tickets and like, you know, update their project managers and so on. So this was the early insight which kind of led us to Merlin. And then while building Merlin, we understood that the scope of what we can achieve with respect to giving context to AI instead of trying to, you know, train AI using your prompts itself. We realized that there is an even larger opportunity. So, you know, that's how we organically kind of moved from, you know, our early developer focused extensions to Merlin and to now Lime.

Utsav Somani: And what do you see? So right now it's packaged as a $19 subscription to many underlying models as well. What do you think, like, in this world of prosumer aggregation play, where is the durable advantage for Merlin going forward?

Pratyush Rai: co-founder - Merlin AI, Thine: So until there is a demand for multiple models and multiple kinds of interfaces to interact with those models, I think something like Merlin does make a lot of sense for a lot of people. Specifically, it makes sense for those category of power users who have an understanding of which model is good for. So maybe you know that, okay, Grok might be good for writing more unhinged social media posts and Claude might be good for more reasoning items and Chargeability might be good for, you know, more creative brainstorming and so on, right? So for those category of users, something like Merlin makes a lot of sense because they can use Merlin to interact with different models and different interfaces, like, for example, a project-like interface or a extension-based interface where you can chat with web pages and so on. So the durability lies in the fact that there is an optionality to the users and we are bringing all that optionality in a very fairly priced package to the end users.

Siddhartha Saxena: co-founder - Merlin AI, Thine: And the point is that there is no single model which is the best across the board for everything. And at the end of the day, for us as a business, it does not matter which model a user uses, right? We are built on the API prices, on the cost. So if you are using Gemini, Anthropic, Cloud, any model, that does not matter to us.

Utsav Somani: Let's talk a little bit about pricing. I've always wondered, how does the gross margin work for an industry like AI where token usage can vary significantly across users as well? So how did you land up at the $19 subscription price?

Pratyush Rai: co-founder - Merlin AI, Thine: So, originally, it was kind of a dark in the dark because we saw that people are acclimatized with paying the chargeability subscription price. And then, later on, what we realized is that people do want the flexibility to be only paying what they actually believe is a fair price, right? So it can be $5, $10, it can be $30. But what does matter for that user at that particular time is that they should feel that the price is fair for them. So even though Merlin today is at a general level price of $19, but for students, it is priced lower at around $5, for businesses, they can go up to $9 to $10 and so on. So depending on different channels, we actually sell different prices in a very targeted manner to those particular kinds of personas who we are targeting with that channel. The second thing is that in terms of gross margins, given the cost reduction curve is just so massive, it is probably the fastest cost reduction in the history of entire tech ecosystem, right? So typically what we have seen is that margins has never ever been a problem in our journey almost in the entire three to four years. Rather, we should have originally kept our margins to be as negative as possible because we did not account for it, account for the fact that the cost reduction is going to go so fast. One of the examples which I always use and which I know a lot of people very unfairly criticized about is emerging. A lot of people talk about that, oh, they are doing this negative gross margin sales and so on. In my opinion, it makes perfect sense. If you are going to have a 200, almost 100x cost reduction in the per unit token price and the use cases are anyways going to always exist, people are always going to want to create portfolio websites or personal websites and so on. So the cost reduction makes a lot of sense to ride on and negative gross margins today makes a lot of sense to ride on, right? Similarly, in our early days with Molen, we realized that...

Utsav Somani: These are not network effects, but the same thing happened in different marketplaces built out as well like Uber and all of these other industries I think were relying on many other network effects or the scale benefits, which I think here is just the token price is going down, right? Exactly.

Siddhartha Saxena: co-founder - Merlin AI, Thine: And it's even true for companies like Claude. So if you know the Claude Max plan, which is sold at $200, essentially, they are giving approximately $4,000 to $5,000 worth of usage. Because the point is that the people who are actually going beyond those limits and are on those plans, they are acting as marketers for the product. They are going, telling their friends, their companies about the product. So essentially, you should not think of it as gross margins of the product, but it's essentially the marketing spend that is going on, right? It is that in the short term and overall, as that visibility improves, those spend will be reduced.

Utsav Somani: And I was looking at a lot of the YC companies. I mean, given that we're doing some forward crystal ball gazing, a lot of the YC companies and even Elon Musk, he's taken a bet that India... Sorry, not India, but India... I mean, on the globe, I think we're going to run out of compute and space and electricity and resources for basically generating... I mean, sort of this intelligence layer that we're creating here. And they want to put intelligence in space now. What is your take on all of this? Like, is this making sense?

Pratyush Rai: co-founder - Merlin AI, Thine: Yeah, it makes perfect sense. We are still not able to digest the pace at which things are improving. Like there is a particular statement which Paul Graham also makes about early stage YC companies that growth rate matters far more than absolute numbers, right? And if you imagine, like today at like hundreds of billions of dollars of valuation and tens of billions of dollars of revenue, the growth at which Entropic and like OpenAI and all these folks are growing, it is as good as probably an early stage YC company, right? Or maybe slightly later stage YC companies.

Utsav Somani: I mean, OpenAI just announced 112 billion round at 850 plus billion valuation. I think Indian VC was 15-20 billion. All of like, I think, I mean, public markets, no single company has ever raised that kind of money as well. No single IPO has raised that kind of money as well. So I want to know what the liquidity looks like. And they also have started planning for this. They've invited some ETFs like ARK Invest and stuff to start inviting retail participation early on just so that they start de-risking their time. I mean, when they enter the public markets as well. So it's just crazy these numbers. I mean, fascinating.

Siddhartha Saxena: co-founder - Merlin AI, Thine: And I think the liquidity is visible in the San Francisco housing market with how expensive like every lot, like house that comes in the market is. Like there are some houses which are like listed for 2.8 million, 3 million dollars and are sold for 4 million dollars. So I think like that just goes on to show the amount of wealth which is being created.

Utsav Somani: I want to get your take on Emergent because that's one company that you brought up recently. They were part of a huge Twitter debate. I think a lot of people are pros, cons, everything. They're on both sides of this coin. Like, what's your story? Where's your support to this whole or what's your view on this whole thing that went down?

Pratyush Rai: co-founder - Merlin AI, Thine: I think there are two parts to it. One part is that as a founder, what are some actually good learnings you can derive from Emergent, right? And as an investor too. I think one of the things is they took a very bold and right call to actually optimize primarily for product and distribution solely in terms of making sure that the experience of the product is good enough. You're not cut short on the free plan. You're actually offered a fair amount of usage. Maybe you are required to go a little bit more negative on gross margin as I was talking about and so on. Ensuring a lot of long tail edge cases are sold specifically for non-technical users like marketers and average internet users. So they have got it right.

Utsav Somani: The second part is The lock-in is also quite high. Like I read an argument how the lock-in is quite high. If somebody is making the website and hosting it with you unlikely that they'll be able to port it out very easily. Exactly.

Pratyush Rai: co-founder - Merlin AI, Thine: So I think they have got a lot of things extremely right which every one of us should learn about. The second thing is that they have probably the best consumer GDM people in the country right now. It does not matter like you know whether they are at a 70 million ARR 100 million ARR how recurrent it is or not. In the history of our country nobody except probably Emergent and NVIDIO has been able to do that kind of a GDM at that scale. And specifically Emergent has been able to do it very quickly. So you know there is chatter in the valley as well that how well they have cracked the influencer marketing and like all those set up right. So I think we should be proud about these areas. The accounting stuff and the revenue stuff I mean they and their investors and a lot of other people including media they are smart enough to understand what the reality is what part of the revenue is recurrent what is a lock-in how to you know include that in the valuation and all the other stuff. So I don't think that there is a lot of legitimate takeaways for founders specifically to be made on those areas. As a builder I think we should just learn on the good things and like on the other parts let them do manage their own business and you know I am super sure that all the people involved including the media the founders their investors they are all super smart to understand which part of what revenue accounts for how much like you know retentiveness and so on. So we should not worry about those things and like just focus on what we can take.

Utsav Somani: There is one of the ex-journalists who put out a post on this and that I think the ending part stayed with me. He has argued for this statement and against this argument as well but I think the ending statement stayed with me that we have still not been able to grasp that a company from India can reach 100 million ERR in such a short time. So I think instead of celebrating that success we always look for ways on how things can go wrong from here. So I think we should just take a step back just enjoy this moment and see where it goes but I think also it's a lesson for founders that I think sometimes it's okay to just go under the radar as well a little bit and not just thump about big numbers very early on because in the world of AI we've seen things can change very very drastically and pretty quickly. So I think word of caution as well on both sides. Let's talk about something that you've launched at CES 2026. I love the website. You're a personal co-founder for life and it's an ambient personal AI that runs on your iPhone and you've called out some other projects by saying that they are kickstarters that got lucky. That statement is pretty cool by the way. So you've decided to skip the hardware bit and go straight to software. So what's the thinking around that? What is the product? Let's introduce it to our listeners.

Pratyush Rai: co-founder - Merlin AI, Thine: So I think the product is fairly simple. It essentially is an iOS app which you can just click on your action button to turn your iOS recording on. So basically it starts recording all your conversations or to turn it off. The app is designed to be something which is on 24x7 by default and you can of course turn it off and manage your data in the way you want. Once it listens to all your conversations it then uses that entire context along with your email context and calendar context to make something, make an AI which is extremely personalized to you. What does this mean? So basically you can use Dyn as a one-stop place where all your thinking and working goes simultaneously. Let's say right before this meet I asked Dyn that like oh I'm going on this amazing podcast and going to talk with Utsav. Can you just tell me what all things should I talk about? So it was able to understand based on my conversations with Siddharth and my other team members what is this conversation going to be about. It also read our email to understand how long it is and to do internet search and identify what kind of themes you like to talk about. And then it gave me different pointers and insights around how I can talk about Dyn. So I used Dyn to actually understand how I should talk about Dyn with Utsav. So this is the beauty of it essentially. It can compile in a lot of different kinds of contexts. The largest part of differentiation and value which we bring to the user is this particular data which is nowhere on the internet and nowhere anywhere else in any other application and that is your in-person conversation data. A lot of hardware products do capture that data in form of pendants and all these things which you can stick behind your phone and all. But as you can think about Like Pocket and Plod I think they're a separate hardware device. Yes, Plod and Pocket. And they have tremendous we have a lot of respect for them because they have proved that the market exists with a fairly poor UX in the form of a hardware device. If you are charging that device carrying it with you all the time and then recording and so on if you do want to use a hardware device to capture your conversations our thesis is that why would you not want to use an app? And the app is not just helping you capture those conversations. The largest part of the value is that it actually then lets you feed into an AI which then works almost like cloud code. Like how you're pulling your different kinds of projects and files and sometimes even using Slack and email integrations inside cloud code to feed it context and then chat with it. Over here, you're feeding your voice email and calendar and having a chat with it. So that is the thesis behind how we are thinking that we want to pull in all the relevant context inside one place so that whenever you're making any decision or planning something later to work you can use it as a one-stop-shop.

Siddhartha Saxena: co-founder - Merlin AI, Thine: And as Vishal had mentioned for example, if they have to make an investment as they're having those conversations as they're having their boardroom meets dining, if they're listening to that they can just ask a question. They don't even need to feed all of the memos and different things that exist because in our conversations we are covering most of those aspects.

Utsav Somani: And this is relying on which LLM model OpenAI or by Anthropic?

Siddhartha Saxena: co-founder - Merlin AI, Thine: So we use a bunch of different models. Again, there are parts of it which are uniquely that we have made out for taking all the audio, transcribing it, dialysing it, doing speaker identification which is an in-house pipeline, taking that, ingesting this into our memory system which uses multiple LLMs. So we use essentially Gemini, Opus and in some parts OpenAI as well. So I then like to get the best of everything and we have multiple evals to just try out which model is improving the performance incrementally.

Utsav Somani: And I think the most important question that you'll always have to answer is privacy, right? So how have you thought about the privacy architecture when building out this whole thing? What are the key steps that you follow to ensure that everything remains safe and there are no incidents in the future?

Siddhartha Saxena: co-founder - Merlin AI, Thine: Yeah. So I think privacy as you mentioned is very important because this data is very sensitive for the users. It is essentially their whole life, right? So something that we have been very clear from day one is that none of the interactions should be accessible to us. We should not be able to let's say see any of the audios that you're transcribing into it or any of the chats that you're having, right? Which in a way is a very common thing but even that's something that we can't touch. Every user's data is also isolated so we are using something called Cloudflare's Durable Objects which means that the compute and the database everything is isolated across users. So there's no scenario even when the data can mix with each other. And then at the end of the day as we go further a plan for monetization will be from the user itself. So this means for everything the data that is coming in is actually to serve themselves, right? None of it is for any other purposes and yeah, that's why again, we obviously have the certifications and everything as well but essentially this ethos is the main part of it.

Utsav Somani: Anything to add, Prathush?

Pratyush Rai: co-founder - Merlin AI, Thine: Yeah, I think the most important thing is that with respect to our journey with Merlin we had to think about privacy even at that point in time because we were building a browser extension taking so much context from browser. So we did think about getting compliant at a SOC to GDPR, ISO and all those levels as a consumer company itself. And with respect to Dyn we are taking it one step further in terms of even just isolating user data and just making sure that any edge cases which we think can go wrong we are kind of trying to contain it.

Utsav Somani: And OpenAI was also coming out with a device as well. Is this going to be I mean, that's a physical device that everyone is expecting for OpenAI to release after they've acquired Johnny Ive's lab. What's the word on the street about that?

Pratyush Rai: co-founder - Merlin AI, Thine: We really respect and love Johnny Ive and that's why we are using the product which he has already made as our primary hardware device. I think it does make sense for a lot of people to attempt to build a phone alternative of sorts in any way. Maybe the interface can be something like what OpenAI is planning in terms of a camera and a recorder and all those things and so on. It's such a large market if you are able to really crack it it's such a large problem statement that it makes sense for people to take a bet on it. What we have recently found to our surprise is that OpenAI is cutting down on a lot of these other consumer projects and at least as per news they are most likely cutting this particular hardware device itself. So the territory for a company like us is now wide open. The only viable alternatives to us right now are using a setup involving something like Cloud and maybe Granola and a few of these other platforms and then integrating with Cloud Code using their MCPs or APIs or whatever and then using Cloud Code to do what you can do with Dyn like you know itself right. So people do make these kind of very elaborate workflows. We have talked to a lot of our users and potential users and understood that this is a very common phenomenon and we want to just completely compress it all that one time you are just giving all access and then it's done. How much are you pricing it at and it's available right now?

Utsav Somani: Where can people sign up?

Pratyush Rai: co-founder - Merlin AI, Thine: We are currently like you know we are currently opening up the wait list we have not actually like ran a you know full-blown price plan because we are also in our early days and this product form factor everything is just super new so we are like trying to just completely perfect it up and you know go through the journey in a similar way how SuperHuman did right like just to make sure that each and every small and large part of interactions are completely perfect because you are going to trust us with such sensitive information and you know such an important part of your day-to-day workflow so we are open in terms of our wait list if you want to join you can just go on dine.com and like sign up on our wait list.

Utsav Somani: Is there a way I can skip the wait list somehow? Yes. I am very happy to directly you know talk about that No, you have got a beautiful website and I saw one of your users post about it on X and that's why I decided to reach out because it sounds like such a cool product because I have some friends who are using Pocket and I mean it just feels too intrusive like it's like a battery pack that's on your phone you need to charge it separately and then there's a separate button if you wanted to listen into calls and not listen into calls so like it's just something that I mean it has to be frictionless so I think what you are building I think is superb and all the best on that journey as well from India to the world so thank you for coming on the show and all the best Thanks Utsav for having us Alright Awesome listeners that's it from us on this Wednesday 1st of April 2026 and we will see you on Friday and Dhruv is returning to this hot seat on Friday Cheers Bye Bye