← back to guests archive

transcript · reviewed NOVEMBER 18, 2026

#episode 155 transcript

Divyanshi Chowdhary

Divyanshi Chowdhary

Kalki Social | OCTOBER 27

Social forecasting platform where people share predictions across politics, markets, business and culture, building a public credibility score through the accuracy of their predictions.

Neel Balar

Neel Balar

Clueso | OCTOBER 27

Y Combinator-backed company that turns screen recordings into polished instructional videos and step-by-step documentation using AI.

Hardik Jagda

Hardik Jagda

Proximity Works | OCTOBER 27

AI engineering company building infrastructure for sports, media and entertainment, focused on putting AI into production for live experiences at scale.

transcript

10,703 words

Full Transcript

Dhruv Sharma: I'm going to be talking about how to get rid of these Today Utsav and I are chatting with Neel Balaj who is the founder of a company called Clueso and these guys are building video infrastructure for the AI era. So Neel, welcome to The Offline Network.

Neel Balaj (Clueso): Glad to be here. Thanks for having me.

Dhruv Sharma: Great. Tell us, give us a very quick version of the founding story. Before we get started, let's look at the TON logo on the bottom left of the screen. Big props to our team that's put this together. Diwali is not that far out and we've got like, and we'll be switching them up too, right Utsav?

Utsav Somani: Yes, that's correct.

Dhruv Sharma: Love it. Love it. Let's get into Diwali more.

Utsav Somani: Awesome. Neel, so we were going to ask you about like the quick, maybe two minute version of the Clueso founding story.

Neel Balaj (Clueso): Yeah, for sure. So I'm Neel. I'm one of the co-founders. My co-founders and I, we were friends at IIT Madras. We all graduated in 2023 and then, you know, during college, we were all friends working on multiple different things. And during COVID, we were actually like super bored. So we're like, okay, let's try and build something together. It'll just be fun. We'll learn some new things. So we started, you know, working on a product together. So actually we were working on a different product before, it was called DeskLamp. It was like a digital reading and note-making tool for students and researchers. So as students, we didn't really remember what we were reading, especially during COVID when we were reading a lot of PDFs. So we were like, okay, let's improve this experience. Let's build a product that helps people, you know, read better, take better notes, and, you know, remember more of what they were reading and also collaborate with their friends and like other researchers essentially. So that's where we started building DeskLamp. Think of it like a, you know, Adobe Acrobat plus Google Docs all in one place. You could read on the left side, you could just highlight something, take notes directly, everything would be linked back and all of that.

Dhruv Sharma: Did you guys at least keep the domain? Because that's a very nice name to have for a company or a product.

Neel Balaj (Clueso): Yeah, we do. The domain is still live. In fact, the product is still live as well. It's just been running on autopilot since Jan 2023. So we worked on it about for one and a half years before we pivoted to Clueso. And the reason why we pivoted was basically with that, you know, we had about 60,000 plus users. A lot of them were using it very actively as well. In fact, a few hours a day, especially legal law students and whatnot. But one thing that we realized was the market was not the one in which we could build a big business. And that's also what YC made us realize. We got into YC during our last semester of college, went to San Francisco. And that's when our partners were like, guys, you've built a great product, but it's not going to work out. It's not going to, you know, this market never pays. And we were initially like, no, we're doing things differently. We'll make it work. Our users absolutely love us. But after two months, we also realized that a lot of user interviews, a lot of conversations with our users as well. And that's when we decided to pivot. We were exploring multiple different ideas, but we actually landed on Clueso because one of the problems that we ourselves faced at DeskLamp was actually we had a great product. In fact, some of the people who were able to figure out how to use it by using it very actively. But there were many others who never even used it. They would just sign up after 10 minutes. They would never come back again. And when we spoke to them, what we realized was they just didn't know how to use the product or utilize some of the marquee features we had as well. And that's when we were like, OK, you know, for this, we used to create demo videos back at DeskLamp. We had created it for specific use cases as well. Those worked really well, but each of them took us about a week to two weeks to make. But like, you know, voiceovers, going to studio, doing all the animations and whatnot. And that's when we're like, OK, let's try and see if we can automate this process or make this process much faster. And that's essentially how we landed on Clueso, which now is like an agentic video creation platform for enterprise teams to create product demos, marketing videos, training videos for customers or internal teams. So that's like a rough, you know, quick story.

Utsav Somani: Awesome. So how is this? I mean, for our listeners, I think it might be useful to know how is this different from, say, like a platform like Higgsfield, InVideo or any others like Rock Imagine that exist. Get a solid prompt, put it there. Will it have the same output or agentic is more iterative in the process?

Neel Balaj (Clueso): Yeah, that's a great question. So all of these platforms are focused on creating videos, which is basically generating new footage from scratch or taking in your existing footage and then trying to edit those videos. Basically, most of these are like generative video models that they're using in the background. With Clueso, what we have been focused on in about 80% of the videos that get created within Clueso are by enterprise SaaS companies who are using it for their product demos. So they do the screen capture with Clueso. So we're focused on videos that are focused for enterprise use cases, wherein you're not generating new footage from scratch. What you're doing is you're giving in your own context or prompt, be it like a screen recording, a PowerPoint, a document. And we help you take that in your own context and turn that into really polished, engaging videos and animations, your branding, a really nice voiceover and all of that. And one of the recent things that we also released, which is an agentic layer for all of this, what is happening is products these days are shipping really, really fast, like five times faster than what it used to be even one year ago, because of which there's a lot of newer features that are being shipped every week, if it's a SaaS product. And a lot of the existing documentation and videos go out of date as well with newer releases. So we have an agent that can now actually navigate through the product whenever something changes, do the updated screen captures for you, create new product demos for new features, update any of the previous videos as well. We're building an entire loop for enterprise teams where it identifies what needs to be created, creates those videos for you, and then keeps them updated as things change as well. And we've focused on specific use cases like product demos, marketing videos, training videos, which involve your product. That's not something that can be generated with AI, which is, I'd say, the major difference between other tools. We still help teams create videos with motion graphics and animations as well. So that's where it coincides a bit with Higgsfield, InVideo, and a few other platforms as well. But the major differentiation is in the context, in terms of the videos that you're building itself, which is the actual screen recording.

Utsav Somani: What are the underlying models that you're using while editing this stuff?

Neel Balaj (Clueso): So in terms of the editing side of things, we've built a fully professional video editor as well, as a web-based video editor. And you can, of course, tweak anything manually if required. But the idea is that AI is doing the majority of the editing for you, and then you can tweak it. And we have built functions just like how there are cloud skills. They're like skills for videos that you can build, but inclusive in terms of the kinds of branding that needs to be added, the editing that needs to be done. And in terms of the underlying model, we're still using, of course, to process any information, transcribe, generate scripts for videos, or even do some screen captures. We're using models like Opus 5.5, GPT-6 Astra, and a few other models like that. But what we have built, our mode is basically the video editing platform that we've built, wherein anything that the AI generates is also something that you can tweak. You have complete control over it. While AI is still automating it, you still have complete control, and you can go in and tweak anything that's required.

Dhruv Sharma: Fantastic. So, Neel, by now we have a good enough sense of what you guys do. What's interesting is all of your customers are also pretty much in the AI race and first-person shooter mode. So as you said, things are moving incredibly fast. I think what will be interesting is for you to give us some examples of how a head of learning and development, how a head of digital marketing, how these different personas within the enterprise have started using Clueso and seen an immediate impact on how they get things done.

Neel Balaj (Clueso): Yeah, for sure. So one of our top customer teams that we target is called Customer Education Team. Basically, their focus is on creating training videos and courses for their product to show their customers how to use their products, for example. So in fact, Atlassian's Customer Education and Sales Enablement teams use Clueso to be able to turn rough and quick, raw, rough screen recordings into really polished videos with their branding, with their guidelines, all of that done automatically for them. So the kind of impact they're seeing is typically what would take about a day to two days to create a video earlier. Now with Clueso, they're able to create it in roughly 30 to 45 minutes for most of the videos that they're creating, which is basically help them iterate faster, create newer content for the newer features they're shipping, but also keep the existing content that they have updated with how fast things are changing as well. So that's how Customer Education Team uses and that's our biggest customer base. Apart from that, we also have many internal learning and development teams or internal change management teams as well who are basically building content for as they're adopting more and more AI, they're also implementing a lot of AI tools on which they need to train their entire enterprise, let's say 10,000 plus employees on how to use those tools, how to use new ERP systems and whatnot. And for those as well, previously, if they were to run those change management programs, it would take them about six months to one year for them to just build the documentation, make sure all the testing is done, and everyone knows how to use those systems before they take those live. But with Clueso, that has come down significantly because all of that documentation, albeit the training videos around it, goes live as soon as the product is live as well or their instance is live, which is something that has helped them compress that timeline to go live with any new product, new process, or rolling out new systems as well.

Utsav Somani: And I mean, looking at your customer list, Atlassian, there's Zendesk, there's GitLab, I mean, Ironclad, there's insane number of names there. Maybe you can share some tips and tricks from your sales playbook on how did you win these logos?

Neel Balaj (Clueso): Yeah, it's quite interesting. It's been a lot of hustle, of course. Early stage, you're just hustling a lot and trying to do whatever you can to get these customers on board. But one thing that I think we've done well is basically once we identified our ICP, which let's say in this case, our customer education teams is the main ICP, because what we realized is while many other teams like marketing teams and other teams need to build videos as well, that's not like the main part of their role as well, maybe 10% to 20% of their role. But for customer education teams, that's majority of their role, which is basically they're creating training content within video format, text format, lessons, or anything else as well. So once we identified this audience, what we tried to do is basically we tried to hit them up from every single angle we could. So obviously, I had personalized LinkedIn messages sending out to all these people, especially to the influencers in the space who are basically very vocal about what they're doing and talking about it on LinkedIn as well. So identified who the most active people are in this space, reached out to them, gave them a quick demo, reached out to them saying, look, we're building this, would love to hear your feedback on what we are building. We're building it specifically for customer education teams and given your experience, that'd be super helpful. And after getting on calls with them, although some of them didn't need it at that point within their organizations, because they're also influencers and like talking about it, they just shared about it on the communities that they were on. Similarly, on LinkedIn and all that. So that really helped. Apart from that, we basically sponsor the podcast that they all listen to. We go to the events that are targeted for these audiences as well. There's LinkedIn reach out that we are doing. We're doing targeted ads and all of that.

Utsav Somani: Will campaigns work anymore like clear tools like Clay and stuff or are they dead?

Neel Balaj (Clueso): I think they do work a lot. Personally for us, that's not been very successful yet. What's been very successful is definitely personalized LinkedIn reach outs and outreach to this audience. And just basically building really good relationships with well-known people in the space. And they definitely help us connect to a lot more people. So email outbound hasn't really been a great channel for us yet.

Dhruv Sharma: You know, Neel, I'm curious. So there's two things that are super interesting here, right? One is just the widespread acceptance of video as the dominant sort of format. I mean, it's just, it explains everything so much better. And then coupled with the fact that you guys can pretty much dynamically update those videos as well. Do you think it's still early to write the obituary of static media content or is there still some time to go? Maybe the other way to think about this is, do you have customers who've doubled down so much in videos that their entire marketing website right now is just a stack of really well-made videos?

Neel Balaj (Clueso): Yeah, that's actually quite interesting and that's definitely the trend that we've started seeing as well. So in fact, many companies that we talk to our customers, we talk to, they told us that they stopped building videos or creating videos because of how time-consuming it is and how difficult it is to maintain them or update them as well. But with Clueso, because now we are helping them create these faster, they're actually, you know, making video a big part of their content strategy overall, be it for training and marketing as well. So while it's definitely increasing, it hasn't necessarily replaced other formats yet. I think every other format has its own place as well, be it visuals, infographics, text-based content as well, because different users have different preferences of following through content or going through content. Some people are reading or watching a video as well. But that said, we are seeing like a lot more videos that are being created by our customers because of how easy it has become now.

Utsav Somani: As a final closing one, you've defined who the ICP is, but have you defined who you're not for? Like before the show started, you were mentioning that you're not there yet for individual creators and their needs right now.

Neel Balaj (Clueso): Are there other like large video markets that you want to capture later? Yeah, so initially we started off with like specific, you know, training teams within organizations to help them create how-to videos, tutorials and training. Now we've expanded to other teams within the organization as well, like marketing and other teams. So I'd say in near term, our focus is definitely going to be on like, you know, B2B companies or enterprises where they're looking to use it for their business use cases rather than, you know, like individuals or like anyone who's using it to looking to use it for their, let's say, Instagram channel or YouTube channel and things like that. The reason why we believe that is because the pricing model changes significantly the way you go to market changes significantly as well. So we are focusing on businesses and enterprises and that's where we believe the bigger gap is as well because tools like, you know, Higgs field, InVideo and a lot of other tools have done a great job for the other audience. But for this business audience, I think there's a lot more gap that we could potentially, you know, solve for.

Utsav Somani: And that's what we are focused on in near term at least. Awesome, Neel. Congrats on the 10 million raise as well. Wishing you the best ahead.

Neel Balaj (Clueso): Awesome. Thanks so much. And thanks for having me.

Utsav Somani: And reach out to us when you do something for a live stream show. Yeah. In the young creator category right now. Clueso for the clip economy is gonna be Clueso, right? So yeah. Yes, that was like this. I mean, a whole bunch of threads which were going on about how they're making skills where Opus and Astra, all of these models are now getting good at creating clips with your brand. So I think it's just take some work. But I hope that the next iteration, they get to a point like automating video editing also.

Neel Balaj (Clueso): Yeah, for sure. That's a goal. We've done that for some use cases already. So definitely gonna expand that as well.

Utsav Somani: All right. We'll be your pilot customer.

Neel Balaj (Clueso): Awesome. Cheers. Thank you so much. Bye.

Utsav Somani: All right, listeners, moving on to our next guest. We've got Divyanshi from Kalki. Divyanshi, welcome to the show.

Divyanshi (Kalki): Hi, Utsav. Hi, Utsav.

Utsav Somani: How are you doing? Fun fact, Divyanshi, when she was at DVC, she was an investor in offline. DVC still is on the cap table. But now she's on the other side of the table. So how does it feel to be a founder now?

Divyanshi (Kalki): It feels, I mean, it was almost everything that you mentioned it would be as you were taking the journey from being an investor to a founder for the offline network. I think very endearing, very lovely. And finally getting to do things that I was observing from the outskirts and looking at people build great things. Now I get to do that. So very much fun.

Utsav Somani: I think that trend is going to increase where a lot of VCs, investors are actually becoming founders. Just, I mean, after bearing on so many exciting journeys, I think they just want to take the plunge themselves when they find an idea. And I think there's abundance of capital right now, especially in specific pockets. So exciting times. But let's introduce Kalki to our listeners. What does it do?

Divyanshi (Kalki): Oh, lovely. So Kalki is an opinion sharing platform. What it started as is that we wanted to build the next generation of social media. And you and I have discussed this before that as AI becomes ever-present, a very integral part of our lives, we are missing on something that is very core to the human nature, which is thinking, sharing, having real world connections. And the way the social networks have been built so far, it's been relying on distribution, right? It's just putting yourself out there. And the only way you really interact with somebody on social is with a like, share or subscribe. That's very limited. You really don't know what people believe or do they think. What is their judgment? So this is an attempt to help them think a little bit more when they're interacting with content. And everybody interacts with content in multiple ways today. So yeah, just a small attempt to make sure that you don't lose something that is very human, which is thinking.

Dhruv Sharma: You know who else shares that view, Divyanshi? It's Elon Musk. There's a tweet from him recently where he says, you're as intelligent as you can predict the future well.

Divyanshi (Kalki): Yeah. Dhruv, that's also very... Sorry, do you want to continue?

Dhruv Sharma: No, no, go ahead. Go for it.

Divyanshi (Kalki): But that's also very simple human nature, right? And from the start of your day to the end of the day, you're making very small predictions. What is the weather going to be like? What is the traffic going to be like? How much work will you get through the day? How energetic will you be by the end of the evening? So it's a very daily part of your life. It's just somehow not coming to mainstream and thinking.

Dhruv Sharma: Yeah, it's interesting. Like we make these predictions on a spectrum, obviously, going from low to high likelihood of the event unfurling in a certain way. And then the job, I guess, is to just minimize the prediction error as we keep going along. But what are the first few months been like? How are customers loving the product? What feedback are you getting from them? Tell us about that as well.

Divyanshi (Kalki): So I think the first experiment that we were running, that we knew this was our belief that opinions, collecting opinions, gathering opinions is important. We didn't know if people will share it in the new format. You're very used to swiping up on any social media, be it X, be it Instagram. And now also so much so on Twitter and WhatsApp. What we were trying to get people to do is actually click on every single card that you interact with on Kalki. And the first thing that we tried cracking was that let's go where eyeballs already exist. So we have two partnerships that are live. One is with NDTV and one is with Entrackr. You all know about both of them. So what we're trying to do is show people that, hey, you've read through, you've consumed some information. How much of that did you retain? What thought did it generate? What is your thought process? So something as basic as if you read about the traffic, were you anticipating this much traffic or not? Two, hey, geo-listed today, were you expecting it to be at this much of a high? These are the posts that are live on both the platforms. And we're seeing a lot of good reactions. We've been actually in the last month, we've been able to collect 150K opinions. And this is just our first proper month of operations and two weeks of live partnerships with both the platforms. And both solid partnerships.

Utsav Somani: But I mean, prediction markets became big because, I mean, Polymarket and Kalshi, you're at least winning something. There's a reward at the end of the thing, right? I'm going to X in posting because either I want reach for the message that I want to put out, hiring post, announcement post, anything. There's likes and bunch of other reward systems. So how do you think about that? Because India with real money gaming anyway cannot happen. No.

Divyanshi (Kalki): So yes, we cannot do anything that is real money. OK, let me take a step back. The way we look at reward as what you also mentioned Utsav is not limiting it into the existing prediction market construct where you put money and then you get certain X out of it. It is actually rewards on how right you were. So when you expand that definition a little bit, which all of our social media does, the reward is sometimes very integral that so many people agreed with what you said, you called it out early and you were able to and essentially prediction markets actually keep 90% has only 10% retention over a period of three months, which means 90% of people lose money and drop out. It does not include a lot of women. It's not very inclusive because it is only for very hardcore people who understand the nature of doing this. So hence they need to continuously keep marketing themselves to bring more and more people. And hence all of us know this, all of us know the news. But even with our past examples of platforms in India who were building this, it was when it was legal, it was limited to male people between the age of 18 to 29 and only on sports. What Kalki does is, it's actually saying that, hey, it's everything. The mechanic remains similar, but why not talk about will like Utsav and Dhruv do the next session with four people? There's so many possibilities that you can gather opinions on rabbit hole content. So I think, yeah, like very, very different from what originally...

Utsav Somani: Who suggests these markets? How do you decide what markets go live? Is it user generated or is it coming from Kalki, like a managed marketplace?

Divyanshi (Kalki): So currently it's us. Kalki is generating content. But as we have gone live with these partnerships, it's moved away from us doing it to NDTV editors and to Entrackr editors, people who know about news, media and production to be doing this for us. So that's something that's very good. Over time, we believe that original content generated by Kalki would be less than 5%. We will be unlocking user generated very soon. As I mentioned, we've just been live for one and a half months, one month properly. So we're very soon open comments. We'll give certain people and influencers options to create their own markets. And as we gather learning from them, we'll expand it to more people.

Dhruv Sharma: So I was trying my hand with the Apple earlier in the day, with the hope that I'll, you know, feature somewhere on the leaderboard. The question I had for you was like, what's the goal right now? Is it to get me to see how close I am to the consensus or actually even take a view on the tail end of probabilities, like the extreme left or right? Not politically speaking, but probabilistically speaking.

Divyanshi (Kalki): Not at all political right now, but there are two experiments that we're running through. One is that, how many reels do you remember that you saw on the same day? But I can definitely talk about that you will remember at least two questions that you read on Kalki. So that's my first experiment. In fact, that was also,

Utsav Somani: I think this is a great alternative for Doomscrollers, like, you know, Matix just raised that around. I think these are incredible places like digital real estate for your eyes to spend time on.

Divyanshi (Kalki): Absolutely. So that's, I think, the first experiment. So you've gotten there, you got to the platform, you were able to see the construct, a new construct come up and you engaged with it. There is a little bit of notion, little bit of human want to be on the leaderboard. We are still experimenting on what it works, but currently what it looks like is that if you have engaged with us enough, you get some Kalki coins. They don't mean anything. They just put you on the leaderboard in a ranking and the top 40% percentile goes through a lucky draw and you can redeem it through an external partner that we have. So we have just onboarded with Hubble. It helps us do some kind of rewards.

Utsav Somani: It looks like a very extended set up to do something. I think red coins ki yaad aagi mirko yeh sunke. I wanted to not make that connect because CRED did something very beautiful and overall then I think. Yeah, the recent digital lottery draw system, I think, helped people burn those points in some useful manner. Correct.

Divyanshi (Kalki): So we're just trying to figure out the economics of where does it not become like CRED that it is exciting, but then it's not exciting. So the relevance of this for me should be always at an all-time high. So that's a very interesting problem statement to solve. But as I mentioned, because we're not looking at rewards, it's just monetary rewards or a redemption that you can do. There'll be certain audience that will always look towards redemption. But as we unlock UGC and give you your position and you know that, hey, you could challenge itself on one of his viewpoints on AI tailwinds directly and you can have a conversation with him on Kalki. I think that's the rewards that we're also going to focus on a lot. So as we do this, we'll figure out a couple of these answers as to what does it look like. We can actually have conversations. It's not just a poll. Like you don't choose one side of the argument. You actually have to like sometimes even defend it with text. So as we're unlocking that, that feature isn't live yet. So give us time till I think end of November. So you will have a comment section also live for you. And by I think December or the end of this year, we will have some huge influencer partnerships like Pin Influencers or Rahul from DVC. I'm already calling him out here. So you have no option anymore.

Utsav Somani: Yeah, exactly. Like people who want to debate. He's super, super active now on X. We've been trying to get him on the show. I think he's coming in December or I think November, right? Oh, lovely. Yeah, he likes to fill his calendar. But how do you think about like, I mean, I mean, when you were an investor, I'm guessing that conversation kept on popping up, right? Consumer social is dead. Consumer social is very hard. All of those phrases keep on repeating themselves year after year. And then you still have new apps that have come up, created new categories. Sometimes they fizzle out. Sometimes they manage to become big businesses. So how do you wrestle with that thought? And how do you design small interactions across the way? Do you have like a consumer psychology person or somebody who's like a solid, solid UX, UI guy who thinks about these small, small touches across the app? Is there an incentive, for instance, for me to invite my friends along right now? So yes, there is.

Divyanshi (Kalki): Whenever we've been, actually, let me answer both the questions. So yes, I think it's been very, very difficult in India to build something that is truly consumer. Something that can match up to what Facebook, Instagram and other platforms like this have been able to create. And the idea is that India is not very deep as a market, but still TikTok was able to get to tier two, tier three parts of India in such a quick fashion because they solved for a form factor which nobody was able to think about, right? So every generation will have something that is very, very new. How do I wrestle with these thoughts that it is difficult? Honestly, by just having blinders on and knowing what your vision is. If your vision is to create some social media that is truly Indian, that is actually differentiated and not just trying to farm attention and cares about opinions. So these are the couple of things that we have been trying to focus on as much as we want to. And with a little bit of money that we have raised, I think we have some confidence that we will get there if we focus in the right direction. What lost a little bit of track in getting one of them to fruition. So that is what we are trying to do. And as I mentioned, collecting opinions, getting some sides and making a reward as a loop while that reward could be monitoring that reward could be just your reach and your presence. As long as I can repeatedly keep proving this loop for even a small set of audience, I think there's some proof point that this work. For Dhruv, your question that, yes, as we start doing more comments, as we start doing more UGC, you would want to show people that, hey, I called it before somebody else. Like I was able to call that Mattix. You productized, see, I told you so. I'm productizing wanna bet and I told you so. Both sides of it. And I think it's very human. Like we were doing this as VCs. We were betting on which companies we believe internally will come out very well or not. And all of them have, we have all been wrong 90% of the times, but the 10% time when we put straight, I told you so and we did. And how did we do that? Very, very loud. Yeah.

Utsav Somani: What is the milestone in terms of number of users where you think like you can actually predict real outcomes? Like, I mean, now poly market, I mean, even during the Brazil elections recently, I mean, the wild swing happened in the last two hours, but that poly market is becoming a good indicator of where the outcome is going to be in the next few hours or next few days. So is that something, I mean, do you have a number in mind, critical threshold that you need to have like 2,000,000 users, 2,000,000 users for you to start accurately becoming a place where, wow, this is what's going to happen? Because a lot of your users might actually be passive also, where I just log in, I don't vote on anything, but I just want to know where the market is headed.

Divyanshi (Kalki): So before we officially started, when we had the Bengal election, so we ran a very small web app experiment to see that without money and with a leaderboard as a concept, do we get to close enough answers? And we did. We got almost very similar results to how counting was happening, even though our polling had closed and we only had 780 users as a base. So we realized that instead of it being statistically large, it needs to be statistically significant. Because we were able to advertise, we knew some people who were campaigning on the ground. We were able to put it through some Bengali influencers in the right crowds. We were able to get to these numbers. And when we tried to do this for other elections and other states going this, we got it horribly wrong, even at the same numbers. So we realize it's not just the number of people, it is also how you have gotten to them. We also ran a campaign on our app on the Delhi University students' election. That was a very similar concept. And there as well, we got all five of the results on predictions correct. And this is again, like we got 2,500 users who came just specific to Delhi University on our platform. And we were able to advertise. So even when I entered, I thought the number has to be in thousands and multitudes of thousands, not like 2,000 or 1,000. That will not give us any results. But fortunately, that was not the case. It was just that you need to get to the right audience, need to make sure that they care about the event, which when you multiply that by number of events, that that is when it starts becoming difficult. And you need to keep finding the right set of people answering the right questions. And that is what we'll figure out. And both the times we got it right.

Dhruv Sharma: I came across a paper where speaking of like the wisdom of the crowds, right? The wisdom of the crowds is more often than not right when it comes to elections. So I mean, on culture, you can bet on some 6,000 different events, but elections is one of the biggest ones. In fact, I think there's $1.6 billion in the line for the US midterms right now. And apparently campaign staff pay attention to the attention of the wisdom of the crowds, because those are like live probability estimates to come back and course correct the campaign as well. I think that might add a statistically significant user basis. And so we're saying I think that becomes a great use case. Imagine Indian election campaigners or something using Kalki to frame their campaign promise. But my question was going to be, I know, I know you guys are experimenting with lots of different categories even now, right? So other than elections and cricket, like what is India genuinely interested in, you know, staking an opinion on?

Divyanshi (Kalki): So when we started doing some experiments, we realized it was pop culture. So people had a lot of curiosity on traitors and big boss and all of these reality TV shows, because these people also have their really hardcore following, right? These are influencers who are going on these shows and their following wants to figure out where do they share their opinion, which is, I think, rather surprising for me because I expected it to be more of like the startup crowds talking about funding and this. And that's where we were advertising. That's what not the case. And in the beginning, but as we have unlocked more channels, more eyeballs, NDTV, it ends up being Indian news. It's about which road construction do they believe? Do they believe the Delhi pollution control measures will actually be effective or not? Will the garbage cleaning that the Gurgaon Municipal is working towards actually work towards something great? So we've seen a lot more of India specific news come in. And as we keep putting Kalki out there to different audiences, I think this will keep getting diverse. And yeah, that's that's where it is right now.

Utsav Somani: Devanshi, what I want to suggest is raise a massive round sponsor to you and the ticker at the bottom. So we need a India version of prediction market. So our space is open, open for sponsorship. But so we can do that without the massive rounds because we are friends. So why don't we get an investor discount? We'll talk about that. Thank you so much for coming on to you and wishing you the best ahead.

Divyanshi (Kalki): Thank you so much. It's really lovely, very thoughtful questions. Please take care.

Utsav Somani: Bye bye. All the best. All right, listeners, moving on to our last guest. We've got Hardik from Proximity Works. Hardik, welcome to T1.

Hardik (Proximity Works): Thank you. And so I'm very happy to be here. Thank you for having me. Hey Dhruv, nice to meet you.

Utsav Somani: You're the best turned out guest we've had like ever. And I mean, so you must definitely not be from the startup ecosystem.

Hardik (Proximity Works): Oh, I am actually. I grew up being I was a software engineer at a multiple multiple consumer startups. I'm currently in London and the weather is like 10 degrees. So I'm stepping out for some meetings and hence I'm like all suited up right now.

Utsav Somani: Explain to us what Proximity Works does as a business.

Hardik (Proximity Works): Yeah, we build a real-time intelligence infrastructure for live sports. So during a live game, our products work behind the scenes and the products are deployed with major broadcasters and teams and they help in increasing fan engagement, in doing the right contextually accurate sponsorship activation. So think of like the engagement, the fan engagement layer and the monetization layer that runs behind the scenes that fans never see. But that's what powers a lot of the live content or live commentary or storylines or like the product placements that happen on the screen when you're watching a game.

Dhruv Sharma: This is super interesting, I mean, I've always visualized this is like a dashboard just sort of lighting up from when the game comes on. So take us sort of behind the scenes, pick any sport or any sporting event and tell us really what goes on behind the scenes. I think it'll just help us understand the work a little better. And yeah, so I'll also cover and that'll also be an interesting start to this next question that I've mentioned.

Hardik (Proximity Works): Yeah, I, you know, naturally I grew up in Mumbai. I started the company out of my bedroom in Mumbai and in India, it's all about cricket. So naturally I started in cricket. I think our first customer was a major global sports broadcaster and the ICC 2019 World Cup was coming up, the IPL 2019 was coming up, 2020 was coming up and it was blowing up at that point. So I was having lunch with the CTO of a major like broadcasting platform and they had a cricket platform and it had become quite legacy over time as they used to support a lot of different types of sports. And I just heard about some problems that they were talking about with the platform that they're anticipating a 3x to 4x increase in traffic. The Indian audience segment itself had so many concurrent users signing up to check the scores and everything. And there's a lot of advertising revenue on the line. So they knew that the platform would not be able to serve that kind of demand at that point. So anyway, I just left the lunch. I didn't have access to the code. I just knew that they had this problem. So I studied the application from the outside, like whatever I could and how it loaded and where the advertising was getting stuck, where the load time was higher and why it was higher. So I spent a few days investigating and I sent them like a 20-page report of findings and recommendations. They hadn't asked me to do that, of course. But they could use the work whether they hired me or somebody else or even ask their own team to work on it. But after reading it, they were pretty impressed with the recommendations and they asked me to take on the rebuild of the whole platform. This is like three months before the 2019 World Cup and there were 30 million users on the platform. So basically an impossible task, but that's where Proximity was born from that impossible task of helping ship that product. And we initially started as a like a services company helping live sports organizations cater to the growing global demand that they are to cater to. But eventually then we started building our own products. And today what we offer is a sports operating system at a very high level. There are three capabilities of that system, which is decide, amplify and monetize. So it decides, it helps sports organizations decide what matters during the live game, amplify what matters to the right audience and monetize what matters through smart sponsorship and activation. I can go even deeper about like a couple of levels if that helps.

Utsav Somani: No, I think this is super helpful, but I mean, so tell us behind the scenes, before AI actually existed, like what were you doing? Was it on prem? Was it on cloud? Like, I mean, how did it work? Because all of these guys, I mean, the latency requirements must be really low because these are real, I mean, fast moving sports, right?

Hardik (Proximity Works): Yeah, I mean, you'd be surprised looking at like every broadcaster works pretty differently. So the ones that we worked with, majority of them were doing cloud-based live streaming, but a lot of the US sports broadcasters are still doing a lot of the traditional, like you have their trucks and trucks are sending it to like the main office and from the main office, you're broadcasting it to multiple linear channels. So it varies from broadcaster to broadcaster. That's one of the caveats of working in live sports as well. There is no standardized like technology infrastructure that every single sports organization uses when it comes to even like the difference between broadcasters, how they use the stack, but also leagues and teams. So like there are some products that are common, but the majority of it works pretty differently.

Utsav Somani: So I'll give you an example like league or team or country or the platform that you worked with in terms of the broadcasting stack. Is it Jio?

Hardik (Proximity Works): It is. It is one of them for sure. I mean, like if you look at the US sports ecosystem, it's just massive, right? Like I think IPL is about $20 billion in like the league valuation right now. But if you go to like NFL or NBA, that's about like $195 billion or $160 billion like NFL versus NBA today. So like in terms of the amount of money that is in live sports, like the US sports ecosystem definitely dominates a lot of that. But when it comes to the fan engagement or the digitization of the whole experience, I think Jio, Disney Plus, Hotstar before that has done an incredible, incredible job there. Congrats to them on the IPO today.

Utsav Somani: Yeah. Are you finishing part of your question answer?

Hardik (Proximity Works): Oh, yeah. I mean, I'll give an example, right? So when a live game is going on, this is one of the workflows that I can go deeper into. So when a live game is going on, you have the commentators who are commenting on the game. But there are a lot of editors who are writing a lot of real time content about the game as well. So anytime a milestone is hit, like so think about like Virat Kohli reaching n number of centuries and breaking like the previous best players record. Now, you know that Virat has reached these many centuries today, like they have those stats based systems that would pop up that this milestone has been achieved. But that milestone has to be converted into a very engaging storyline. In order to convert that into an engaging storyline, you have to do a lot of historical comparisons with prior major players, with other teams that Virat has played against, with which other stadiums has Virat had centuries and what are the weather conditions of those stadiums? All of these are different types of variables that are taken into account when you write a very compelling story. Before AI, a lot of that work used to happen manually. So we had these machine learning based systems that would tell you that this stat has happened. But then after that milestone is achieved, you have to convert that stat or milestone into an engaging storyline. And all that work used to happen manually by like an army of editors and stats guys who would sit behind the scenes from the commentary desk and who would go through all of these comparisons entity by entity. So it's like a player by player comparison, a team by team comparison or the stadium comparison itself. And they would take about 15 to 16 minutes to actually craft the perfect story. But by that time, the moment is already passed, right? So you can't really use that real time story in the commentary for the commentary desk, but you can still use it for like different social media platforms or any other places where you're publishing your content. But the live moment has passed. So how do you, one of the systems that we built was how do you take advantage of the real time moment or milestone and craft that perfect story within a minute. So that was a target that we were working with. Like, can you craft that story under 60 seconds so that it can reach the commentary desk in time? And that's what we did. So we had like what we built, which became our core product was we have a team of AI agents who are basically monitoring the game as it's happening. It's anytime a milestone is hit, like there are about 12 to 14 agents who are doing a lot of these historical comparisons and fact checking. So some of them would do the comparisons. Some of them would work on double checking the comparisons. And the last one would do the eval and the check of what the result is and make sure that the story is 100% accurate and can be published. And then that story would pop up on the screen and people would use that story. The commentary desk would use that story. The editors would use that story. And even the marketing teams and commercial teams would use that story depending on what that story is all about. So if it's a Virat Kohli story, you can use that to not only talk about Virat's performance and the team's performance, but also who are the sponsors in that moment who are contextually relevant? What are the products that Virat is endorsing? Can you use that product placement at that same moment in time? Because when you're seeing a product which is contextually relevant, you don't really get bothered by it. It's not interruption advertising and it just flows into the whole experience. So yeah, that's one of the behind-the-scenes examples.

Dhruv Sharma: You know, I think what I'm getting curious about is when these live streaming platforms and broadcasters bid against each other to win exclusive rights to show a certain event, does the quality of their technical team and recent experience actually extend them a durable advantage in winning that bid? Then Zee5 shouldn't get anything like this.

Hardik (Proximity Works): I mean, some sports in some regions are of course political, but... I think there is definitely a variable or that's a point to consider when they are getting those rights awarded. And especially if there are a lot of public complaints about the platform, for sure. I think that is a consideration. But in the end, it's also... There are so many different channels that you are streaming the game on, or it could be linear TV, it could be streaming channels. And a lot of it also matters in terms of how much you can monetize that. Because for the leagues, it doesn't matter whether it's cricket or any other league. The point is, like, they are selling the rights at a certain valuation and the broadcaster needs to be able to make that money back. So whether they give technology that kind of consideration or not internally, it's about a business decision more than anything else.

Utsav Somani: My brother-in-law has marketing and partnership for the cricket league in Dubai, ILT20. I happened to pass by with him in different rooms where the set up, the broadcaster and team owners and all kinds of different things were happening. But they've never mentioned a platform like this. Have you pitched to them? Is it useful to them? How does it become useful to a league? How do you price it? Like, is it per insight? Is it per hour of footage process? How do you pitch to a league?

Hardik (Proximity Works): Yeah, we've not pitched it to them yet. My core focus, like, over the last few years has been just focusing on the US sports ecosystem. We are now actively... I started in India and, like, global cricket first. So a lot of those discussions are still in the pipeline, but we are not, you know, we've not really reached there yet. My focus has been on, like, going live with basketball, hockey and soccer, which is football. So those are the three sports that we're looking at right now. But in terms of how we deploy this platform, it's, like I said, the technology is very fragmented. So it's not that you can just plug the platform, it starts working. There's a fair amount of forward deployed engineering work that goes into actually integrating everything with their current systems. And we can't force a team to use our platform to do something that they do on a daily basis. You need to be where they are. So technology needs to be the fabric there, right? So you need to make sure that the way your platform gives insights or the way that your platform works, it's integrated into the tools that they use on a day-to-day basis instead of forcing them to change their behavior and then, you know, use a new screen or a new platform entirely. So like it's, yeah, I'd say depending on how complex the system is, it takes anywhere from one to three months at least to finish the integration, do the forward deployed work and have them use the platform. In terms of how we price it, it varies from sport to sport, but most of it is usually charged by the hours of game feed that we process and the number of insights or the number of modules that they want to activate on that OS. So we have like an engagement module that writes the storylines. We have a sponsor activation module that activates the right sponsor at the right time and that is a difference. So each of these modules have like a fixed annual fee and then there's a usage-based fee that goes on top of that. Arthur,

Dhruv Sharma: Hardik, can sports fans look forward to an immersive viewing experience anytime in the next two, three years? Is it already happening? Have I been living under a rock? Yeah, it has started happening. What do they do now? The Vision Pro F1 partnership, I think Apple had a partnership with F1, I remember, the Vision Pro thing, although it's very bulky to wear, but you could literally like sit in the middle of a track, middle of a car, like while the actual live race is happening.

Utsav Somani: You can literally see your room I mean, it's fascinating what that experience is like. And so much more fun than just computer screens and phone screens, sitting at a bar and watching it on a projector. But yeah, yeah.

Dhruv Sharma: Hardik, anything you've come across in addition to that?

Hardik (Proximity Works): Oh, there are so many of them. Like there's a Cosm in Los Angeles, there's the Sphere in Vegas. So if you go to any of these places, like they have these massive shared reality experiences where you are in that immersive experience, but it's also not just about, you know, watching a lot of the people pay a premium to actually go to these experiences instead of the real game as well.

Dhruv Sharma: Can you go get tackled by a 300 pound linebacker if you wanted to, if you're feeling particularly adventurous?

Hardik (Proximity Works): Sorry, sorry, could you repeat that? I'm saying if you're feeling very adventurous, can you get tackled by a 300 pound linebacker like this in Vegas or something? I have not come across that particular use case yet, though. But I'm sure it's possible, at least in the glasses, I'm sure it's possible. Like, I think Utsav gave a great example about the Apple Vision Pro, even the Apple iPhone. I think when Apple got the rights for F1, I don't remember, it was, I mean, I think six months ago or something, I was watching the game. I'm watching the movie, actually, the F1 Brad Pitt movie. And like you could actually feel the, like when the cars were taking a turn and whatnot, and if you're watching it on Apple TV app on the iPhone, you could actually feel the haptic feedback with the cars. Like that's the beauty of our Apple's ecosystem, right? Because it's so tightly integrated with hardware and software that they can actually have those, you know, those small pixels on the screen convert into like feedback on the phone and like all of that, that happens. So I think there is, it's getting there, but it's getting from market to market. And I think it also depends a lot on accessibility. So the tickets for like certain games in the US are extremely expensive. And for you to be able to get them, you'll be shelling out thousands of dollars, but you would go to this, one of these shared reality experiences and you can get the tickets for a few hundred dollars. And also the other thing I was telling you about, like a lot of people prefer doing this is because you can also go with a small group of your friends instead of you being forced to watch it with like other 50, 80,000 people, right? So even with FIFA, like when the FIFA World Cup was happening in the US, like 50, 80,000 people, I think got in the stadium, but there were hundreds of thousands of like watch parties that were happening from every single like apartment complex and building to every bar. And each of them had their own sort of ecosystem that they were powering. Now that was not like any kind of immersive experience, but it was more about just watching the game that you love with the people that you want to watch it with.

Utsav Somani: If you want a real immersive experience, then just go to Odyssey in like 4DX. This thing is like water. That was a true OG, authentic immersive experience powered in India. But I think this was fun. Thank you so much for coming on our show and hopefully we'll connect again soon with you.

Hardik (Proximity Works): Thank you, Utsav. Thank you for having me. Thank you, Dhruv.

Dhruv Sharma: Great to have you. I just hope that water gets filtered.

Utsav Somani: Man, can't, can't, this thing, I mean, I think in Noida, so yeah, can't predict if they're actually, yeah. But anyway, we have to talk about the Anthropic IPO today. There are actually two IPOs that have happened. One is going to the public markets, Anthropic, and one is the Jio Platforms IPO. You saw any of the clips coming out of the NSE bell ringing ceremony?

Dhruv Sharma: I did, I did. That's happened.

Utsav Somani: Yeah, that's happened today. Top of the pop, 21%. Yeah, people who go to Lotred are up by that much. And second most valuable company in India. So what a milestone.

Dhruv Sharma: I think second most valuable after Reliance, right? After Reliance, maybe top two.

Utsav Somani: More to them. Let's talk about Anthropic. It's going public. So 16.3 billion in six months and an 80.7 billion loss that is mostly paper. So let's do a quick readout. On Monday, Anthropic made its IPO prospectus public. On 26th October, to list under Nasdaq, ticker ANT, with Goldman Sachs and Morgan Stanley leading a 22-bank syndicate. The headline loss number, which I just mentioned, is slightly misleading because 80.7 billion means not all of it is cash. It's basically 72.4 billion non-cash charge on convertible notes. The business itself, in Q2 of this year, has an operating cash flow profit of 974 million, 974 million. So I think this is the first time anyone is getting to first see what a frontier lab economics really look like properly, without any reported leakages and all of that stuff. And revenue growth is not the question. I think compute bill, governance, many other things are coming to light for the first time. And I mean, it's a fascinating read. And Patrick has done a great job at it. Dhruv, do you want to give me the highlights?

Dhruv Sharma: Yeah, I'm just reading them along with you, Utsav. But I think, again, Patrick has a good distillation of how the revenue ramped up over a period of time. What's the split between enterprise and the consumer business? Average revenue per employee is the startling figure over there. And the fact that the amount they spend on compute has consistently been coming down. Of course, they still spend more on tokens than they do on their own people payroll. But that's to be expected from a frontier lab. And then, of course, there are some accounting adjustments and so on. I think that is, that's like the high level summary.

Utsav Somani: Do we, would we know when they actually go ahead and do the listing? I have a feeling of first week November. Yeah, people were, there were a lot of people who were expecting OpenAI to drop their S1 first. But well, at least OpenAI is doing apparently a private round 30 billion at 1.4 trillion. This one is targeting to hit the markets at 2 trillion valuation, man. Like these are like five years old. But yeah, companies, not even five years old, actually. I think it's 21, 69 months, actually. Yeah, yeah. But the revenue concentration is very crazy here. I don't know the exact stats, but I read, I don't know if we have this here, but apparently like some crazy amount of revenue comes from literally, like, I think, 20 names.

Dhruv Sharma: Yeah. A big, big part of revenue comes from very limited set of names. I'm pretty sure they disclaimed it as a concentration risk.

Utsav Somani: Concentration risk, but they also pay out a lot to SpaceX. So Elon Musk, of course, is a winner here. I think they're paying almost 3 billion a month when the peak projections hit. And both of them can walk away from this deal with a very short office. Elon, I mean, it's still news already, given how the air hype cycle moves, but Elon had, I think he just popped one of the air bubbles like a couple of days ago. I think it was something to do with the New York Cloud core weave story, but I think it's already.