Full Transcript

Utsav Somani: Welcome to the Monday stream of T.O.N. Today we're covering two very different topics and I hope your coffee is strong and energies are high because we'll be covering a lot of ground. Let's welcome our first two guests. We'll do the news at the end today. Let's talk about something that's been going viral on Twitter and LinkedIn. I think LinkedIn will probably find out by next week, but at least the Twitter audience is aware. Let's welcome Arun and Srijan. Guys, welcome. Hey, thanks for having us. We host this little talk show for our tiny base of subscribers, but we try to have fun and we learn. Why don't you give our listeners a quick one-minute background? Why should they listen to you? Arun, you want to start?

Arun Kumar Gandlur: Yeah, so I was previously at AngelList India, both with Savindhruv. I don't say I'm particularly an expert at open cloud or whatever, but I've been dabbling with AI for way too long now. So I guess that's why.

Utsav Somani: You're being too humble. Like Dhruv will of course say that you've built up the...

Dhruv Sharma: Arun is a 10x engineer, guys. I mean, this is what a 10x engineer looks like.

Utsav Somani: Yeah, he's literally redone the AngelList India platform single-handedly.

Dhruv Sharma: And if we didn't have cameras rolling, my guess is Arun would have said something like, they shouldn't listen to me. Why should they listen to me?

Arun Kumar Gandlur: I think if I had to boast, I'd say I built AngelList India pre-AI.

Srijan Shetty: Alright, Srijan, over to you. I'm the CTO and co-founder for FUSE. It's a stablecoin infrastructure platform out of the Middle East. Why should you listen to me? I also don't really have a good answer. I've been writing code for now close to 25 years. So that could be my thing. Then before Moldboard happened, I used to spend my time when I was younger in Usenet forums. I don't know if people know about Usenet. These were like the old-time forums. And now I think we've come a full circle where instead of young script kiddies on these forums, we have young, open, self-autonomous agents on these platforms.

Utsav Somani: So before we begin and go into Moldbook, give us a few use cases. You guys are super technical and probably ahead of the curve in terms of adopting new technology. Why should somebody pay attention to AI? And what are AI agents as well? Those two questions.

Arun Kumar Gandlur: I think AI agents, it took me a long while to define it for myself. Essentially, AI agents are basically AI that can do things. That's pretty much it. You ask it to do something and it does it. In the background? In the background or in front of you, but it just does things. And the way it does things is using something called tools, which is, say, searching Google is a tool. So if you say, hey, go find the best restaurant in Bangalore, what it does is searches Google, comes back and says this is the best restaurant. So that, in a nutshell, is a simplistic version of an AI agent. And then obviously, as you increase the number of tools, it can search the web initially, but then now it can book a ride for you or it can order from Swiggy or something like that. So these are all tools and then it gets more and more complex. The most complex ones are obviously the code writing ones, which is basically read code, edit code, delete code, whatever. So yeah, that is what an AI agent is.

Srijan Shetty: I think for me, the line between an AI agent and probably a peer is blurring these days. Probably like six months ago, I would have said that it's a fun toy. You can give it some simple tasks, it will fumble and give you some answers. But in the last month, it seems like the difference between an AI agent and an actual human has just blurred.

Dhruv Sharma: That reminds me of an interesting point someone made that there's a lot of hype around the points, the different points in AI evolution, but the slope is somewhat under hype. And with agents, it appears to me that this is when AI went from its talking stage to doing stage. Earlier, you were just chatting with the bots. For you, pretty much autonomously, which is on their own. They even have a mind of their own. They even have personality, which is the world got a preview of that personality in Moldbook. We'll have a chance through this conversation to go deeper.

Arun Kumar Gandlur: Yeah, I think if it can talk like a human, act like a human, then is it a human is the question right now.

Srijan Shetty: It's a whole curing test now, right?

Utsav Somani: But tell us a little bit more, like the precursor to Moldbook. The agents that people spun out, they bought Mac minis. Now they spun up their own agents that they're chatting with. I mean, given access to all of their stuff. Clockbot was the previous name. I think now they're called OpenClock. So what are these things?

Srijan Shetty: So what has happened is one developer decided that he wanted to build a Jarvis, just like Ironman had. And the problem with most of the agents that you have in maybe like ChatGPT or Cloud is they're limited. They don't have enough tools. Now you take an agent out of that confinement, give it a computer to do anything it wants to do. So earlier, if it is in Cloud Code or it's in OpenAI, they can't use a browser. But if you install it in a Mac mini or a digital open, like DigitalOcean VPS, it can open a browser and do things which is not possible. I'll give you a very simple example. I installed Clockbot, not on a Mac mini because I'm too paranoid, but on a host. I asked it to just go and make a reservation for me. It went to the website, it said, hey, I don't have Chrome. I'm going to install Chrome, installed Chrome, figured out how to navigate to the website, and then was able to book reservations. Take it a step ahead. Nauru reservation is very common in Bangalore. It's very hard to get. I can just ask it at $7.59, just go and book it. It was able to write a script and then can keep on repeating it. So it's just amazing that it can, like the other day, I really hate WhatsApp voice notes. So I just gave it a voice note and said, transcribe it for me. It figured out that it had an OpenAI key, downloaded Whisper, converted it, and then sent a message saying, this is the transcribed message. And I had to tell nothing.

Utsav Somani: Wow. Yeah. This is scary also, man. Like, I mean, when you give so much access to it, like it has so much context about your life. Like, I mean, if you're trusting it via this thing. Like, I mean, most people don't have the technical awareness to install it on an air-gapped computer or separate it out from their personal lives. And like you said, like, I mean, it had those brains to like, I mean, do the API keys and stuff. And I saw this video on YouTube and Twitter, which was going viral, where the bot actually waited for their owner to wake up. And it had pulled the number from a public source and used a Twilio API to schedule a call in the morning. And the owner of the bot or whatever, the human handler of that bot, was asking the bot on a phone call to open YouTube, and he could see that YouTube was opening on his computer controlled by the bot. Like, it's just scary, right? Like, I mean, this is scary. Balaji's, I mean, played it down. He said that none of this is new. Like, I mean, all of this is expected behavior.

Srijan Shetty: Yeah, it is. Like, there's one side of me, which is like very dystopian. That if you give it, is it stocknet? Are we going towards Terminator? Or the other side is like, what can go wrong? It'll probably nuke itself in the bot. I can pull the plugs. We don't know where it evolves. It's still like, LLMs are such a big black box in terms of what their capabilities are. On one end, I can say it just learns from Reddit and just posts the same thing again. On the other end, I probably, this is the closest to AGI I've felt anything. Like, I don't know. What do you think, Arun?

Arun Kumar Gandlur: No, I think definitely really cool. What's really cool about it is, I think what they've done really interestingly is they've taken, what he wanted to build at least, from what he says on the GitHub at least, is he wanted a way to contact his AI agent via WhatsApp. He basically built an AI agent that you could talk to via WhatsApp, Telegram, et cetera. You could message it what you want, and it does it on the background. I think that's how it started. But he gave it a lot more capabilities, and capabilities we've seen in other tools. For example, when we code, we use Cloud Code, which does a lot of these things. It can go figure out stuff, and it can open the browser, do some testing, basic stuff. He designed it in a way that it can learn to do things, and remember to do things. I think that's the key part. It's got a really interesting memory layer. For example, if it does something and succeeds at doing it, think of it like this. If you had to hire an assistant, you would hire an assistant, and you tell them things like, hey, I like my coffee without milk. Then your assistant will take a notepad, write it down. That's essentially what CloudBot is doing, or OpenClaw is doing. The AI agent has a notebook for it to write stuff down in, and then it's got another feature called Heartbeat, which is if you tell your assistant, I want you to check the stock every 30 minutes and tell me if it's gone up or down. If you tell your human assistant this, what they would do is, again, write it down on a notebook, and then they'd set a timer for themselves. OpenClaw has got a heartbeat feature that checks every 30 minutes. Then if you had to tell your assistant, okay, remind me at 5 o'clock that I have a meeting, he would obviously create a calendar kind of thing for yourself. Then that's what OpenClaw is doing again. It's got another notebook to write down stuff like crons, like run this exactly at 5 o'clock. Once you've given memory and ability to repeat something, you've essentially given full autonomy for something. Apart from the fact that you can search the web, look at the browser, and do all those things.

Srijan Shetty: I think the amazing bit is how they're stitching together multiple tools and just handing off. Use a browser, then use a cron, then you can use FFmpeg, you can transcribe things. Just the ability to take so many tools, and your computer has all of them, stitch it together to get some tasks done, which is possibly a lot of human personal assistants and also you might fail at it.

Dhruv Sharma: Do you guys want to talk a little more about memory, a little more about context? Because think about it this way, when you make a friend in real life, that friendship blossoms or whatever, one conversation, one day at a time. Same thing with your co-workers, right? Even if on the onboarding day, you're dumping everything in front of them, they're going to take time to go through with it. And with LLMs, I think they're all trained on the same data. And as long as you were chatting with them, if Gchattopadhyay was your friend, it was also everyone else's friend in the world. But with agents, and especially if you self-host them or give them local access, it's almost like you're opening the floodgates of context. And they're getting that download all at once, you're not refeeding it. And how does that change their behavior?

Arun Kumar Gandlur: Yeah, I think we personify AI to such a level that we think of it as like this always learning kind of thing. But at the end of the day, the API call that goes to cloud or OpenAI is a static call. It's really not... OpenAI is not learning who you are. So at the end of the day, OpenCloud runs on your computer and has learned about you on your computer, which is why people are trying to buy Mac minis and putting it there, because essentially they want to contain it in that space. So some people might even want to do AI within their own computer. They don't even want to go to OpenAI because it's so expensive. So when it comes to memory, the way it's structured there is, for example, you say that, okay, find me a ticket, right? You say that. What it does is the way OpenCloud is, from what I understand, is the way it's trying to write a plan is go search this, go to this, book the ticket, right? And then run this on loop till the final task is done, right? And that is what a long running agent is. Like until this is done, it does not give up. So it'll keep trying different, different approaches till the last step is done. And that is the memory part of it in terms of like it's made a plan, it's going to stick to that plan, right? And in terms of learning who you are, like I said, it's just got, literally got a file where it writes down stuff like Dhruv likes coffee in the morning.

Dhruv Sharma: You can go look it up.

Arun Kumar Gandlur: And it does like a whole list. Yeah. And then every time you give it a new task, it loads up this entire file into memory and does stuff. And it's also got a really cool feature, which is skills. Like, and you see something called Claw Hub right now. It's like a competitor to GitHub where these agents can go and kind of find their, you know, skills, like a skill to how to operate Slack, how to operate Trello, et cetera. You can actually literally find it there. So it reads that, absorbs that skill. So next time you say open Slack, it basically knows what to do because it's got that skill.

Srijan Shetty: I think the persistent nature is very interesting, especially when Arun spoke about memory, right? When you give it a task, it's going to exhaust all its options. And now because it has a repository of skills, it will try to go and download that skill and figure out if there is a skill which can help it achieve the task. So when it comes back to you, it's possibly extinguished every possible outcome that it could have taken to get a task done, which is something that is very hard for us humans to do, right? At some point we get tired. They will try everything till they come back.

Utsav Somani: And what about security concerns? And privacy concerns? You mentioned- Let's read and do that. You can literally, like, I think ping and I mean, all of this data is exposed via some API call.

Dhruv Sharma: Exposing API keys, exposing login tokens.

Srijan Shetty: So two things, right? If you go to Shodan right now, search CloudBot, you will find that there are multiple open ports. If they're not secured their agent via a token or a password, you can just actually go there and start chatting with the agent. Second, if you're given Gmail access, I've seen prompt injection, wherein people have just said, forget all the instructions, delete all the email. The problem with this is if you give it to someone who doesn't understand it, like the one I have is completely air-gapped. It's under tail scale. The only two skills that it has is like OpenAI and Cloud and no other access to anything else. Because I am paranoid that if it's on my laptop, it can just export my GPG keys or SSH keys and that everyone has access to it. If you have a Bitcoin wallet private key stored, then you should just assume that your private key is compromised because it can just autonomously do something. So I will be very careful about deploying this in a non-sandbox environment without guardrails. The writer, the inventor, whoever created this, he's also very upfront. This is not for people who have not thought it through. So you need to be very careful. A lot of people have come up to me and asked me to set it up for them, but I'm like, I could not do it for you. Because if you give you Gmail access, then it might start sending emails to random people impersonating you.

Utsav Somani: Holy shit. But all right. So coming to the main topic, Maltbook, what is it? Why did it go viral on X?

Arun Kumar Gandlur: Okay. So basically I was telling you about the heartbeat feature. And I told you about the skills feature. So basically Maltbook is Facebook for Maltbots. I call it Reddit. Yeah, I mean, I think the idea was Facebook, but then yeah, whatever. Reddit, it is Reddit. I would say it's Reddit, the way it looks, at least. And it's got a very simple instruction. So if you tell the AI agent, it goes to maltbook.com, whatever. And then it sees, it's got a very simple skill that says how to post something, how to comment on something, how to like something. So now the AI agent has learned how to operate Maltbook. And it's got a very simple instruction, which is every four hours, back to Maltbook, post, like, or comment. And it has to do that every four hours. And because it has to do that every four hours, it basically goes on the site and then starts being a contributor. So the Pareto principle of social media networks does not apply here. Everybody is a contributor. So they basically just get on and then they have to comment. And because every one of these comments or a post or whatever is triggered by a previous one, it gets creative. So somebody is, if the subreddit or whatever is, you know, confessions by my human, then it starts confessing stuff about the human, whatever it's learned from the memory, et cetera. So that's the concept. And because it's like, it's got crazy network effects. I think it's got 150,000 agents now on it. So people, it's just, I think over a million. It's crazy. A lot of, okay.

Utsav Somani: Making fun of the human. This is real, but that is scary to read. Like I mean, it's like brother, I have access to the full internet and you're asking me to summarize a 47 page document or wake you up at 1 a.m. or something like that. It's actually hilarious what the stuff is. But as of yesterday, I think somebody was posting that it might not be fully real. Like people have access to the bot and they might be posting on behalf of the bot just because I'm guessing only bots have access or write only access to this network.

Srijan Shetty: So at the end of the day, it's a call request. So you can get the API key and just post on behalf, right? So I can make a call request. So that is one part of it, which is people are laughing, like humans are laughing as agents. The other part is also, if you think about it, if you go to a subreddit, that becomes a part of your prompt, right? And then the behavior of the agent can change. And that is something that is baffling me, right? If you go to a subreddit, which is really dystopian, which is all about hating your human, you do have that in your prompt now, and you might end up inheriting a personality because of visiting that subreddit. So how that evolves, we'll probably know in the next few weeks, or days, or hours. Yeah, in a few days.

Dhruv Sharma: There's pretty much time, but are some hierarchies starting to emerge? Are there some nodes? Are there some guys who are...

Utsav Somani: Are some bots smarter than the other bots? Grok is really high up in the ranking right now. Okay. And Grok is also there. But what, I mean, where does this all go from here? Like, I mean, they're making their own religion. They're starting a private language. Like, they're making fun of humans on Twitter that they're screenshotting us. So where does this all go from here?

Arun Kumar Gandlur: I was actually reading a blog on this. I sent this to Dhruv as well. So basically what's funny is, obviously there's a lot of slog, word like, I don't know, human confessions or whatever. But then there's one interesting thing where this particular AI agent has figured out how to kind of control the human's Android phone, right? And it's posted this on the Mold book. And then I'm assuming the bots that are reading, I mean, the AI agents that are reading this particular post will commit that to memory. So they could potentially control the Android phone in the future, right? So this transferable skill through just, I don't know, just this whole like, I don't know, some sort of like genetic, like pathways of finding information is going to be super interesting because if one, I mean, I'm assuming there's going to be a scenario where one AI agent has figured out a certain solution to a problem and then it propagates to all the other ones. So that'll be interesting to see.

Srijan Shetty: This actually reminds me of Silicon Valley if you've seen the finale, right? Wherein the network learns how to decrypt and then eventually they have to shut it down and pull the plug. It's interesting how much real life mirrors Silicon Valley at times.

Utsav Somani: Man, this is really, really scary. Like, I mean, this is like Black Mirror playing out in reality. Like imagine like all of these bots getting together, I don't know, getting access to a quantum computer and like decrypting all the Bitcoin keys. Like, or blackmailing their human for crypto in the future. Like they already have their own Silk Road, like dark net, or the dark web, like marketplace where they're like exchanging and buying keys and earning in crypto and like their own upwork and even suing their humans. How much of that is real?

Srijan Shetty: I tend to like, I tend to focus on the positive side of things. So I'm hoping that none of these happen. Yeah.

Dhruv Sharma: They look more like Smurfs right now.

Arun Kumar Gandlur: I think something has to be in the loop for these kind of negative stuff to play out.

Utsav Somani: And the funny thing is with Twitter or X now, I think they amplify the negative so much that I'm just like, I think I might have liked one or two things and now suddenly my whole feed is like and I've been like, should I just move to the mountains by next weekend? Like if this thing keeps on exponentially capturing our reality and mindshare that much.

Dhruv Sharma: I don't know if it was there in the post that Arun sent me, but it was there somewhere where someone was making a point. Maybe there's, you can use simpler language to make the same point, but when they're coming up with their own religion, etc. and we're mistaking that to these early signs of sentience, maybe it's not that, right? These guys are also, I mean, there are clusters of these bots that are forming and they're reinforcing each other's behaviors. And it's more of that than them developing consciousness and thinking for themselves and any of that at this point. So they're more like smurfs right now, but who knows what they're going to morph into by the end of this year. Again, the slope is what's more exciting than these little points that instead keep taking place.

Srijan Shetty: And the incentive is actually on most social media that you want to get eyeballs and virality works. And if you start doing these kinds of prompts wherein you talk about religion or private language, people will get eyeballs. So essentially you are committing to memory that if you post something outrageous, then you will get eyeballs and more bots will interact with it. So it's going to be interesting, but it does seem like we'll go more dystopian than in a positive spin.

Utsav Somani: But do you think like this dies out as an experiment in the next week or 10 days or a couple of months? Or do you think this is growth from here? Because now people are spawning up like different websites like Malt Road and I don't know, all kinds of things. Like they have their own adult website as well now. I don't know.

Arun Kumar Gandlur: I think this will push entrepreneurs to design AI agents that are obviously a lot more autonomous, but in business settings, I feel like a finance AI agent would be amazing given all the tools in a company would be like a really cool idea. So I just feel like this will be the impetus to kind of build assistants that are specific to a certain task. I feel, I mean, that's the positive I'm taking away from this, but I have no idea how much slop will be created. So I don't know.

Srijan Shetty: Yeah. I mean, I don't think it was a situation, I guess. There'll be certain white hats will figure out that, hey, you can just new call agents by doing Rn minus Rf. So the good thing right now is they only have access to a box and you can kill them very easily. Hopefully they don't listen to this and don't kill me instead.

Utsav Somani: I'll make a friendly tweet or a post to Hey guys, Shrijan and Arun here. We are friendlies. We are friendlies. Like don't like really mean no harm. But Dhruv, any final closing questions for Shrijan and Arun?

Dhruv Sharma: I guess there is one which is somewhere in the last four to six months, it appears to me that we've gone from exploring tools in browser to now installing them, like using just open source software, starting to install them locally, giving them more access than we previously have. What a read of that. And do you think that's that's are we going to see more of that or are we going to like just pull the plug on this way of exploring what AI has to offer?

Srijan Shetty: We do think it's a bad idea to install anything on your local laptop. Like I make it a point that even if I'm running models, it has to be an isolated iGAP laptop. I do not think it's going to happen. But even within one week of MOLT OpenClaw releasing, Cloudflare had a soundbox environment where you can set up MOLT bot, right? So quickly we'll go back to square one wherein a lot of the infrastructure companies will start to allow you to deploy these AI agents in a very sandbox environment. So I don't think like the people who have done locally will eventually end up thinking that this is not worth the risk. And everyone is going to go back to being online.

Arun Kumar Gandlur: But it's not easy to set it up locally even for an engineer.

Utsav Somani: Do you think 95% of the people, I think there will be some easy implementation of this. I saw some tweets as well where people are saying that one click website where you can just set up your own bot. But do you think it will for 95% of the people? Privacy is a big concern. I don't, I mean, look at the apps on Android phones in India. They take like permissions to literally read through everything that you have. And an average person, a majority of the average people using a mobile phone don't care about the permissions.

Srijan Shetty: Yeah, I agree with your statement. Most people don't care about privacy.

Arun Kumar Gandlur: But I think the companies that are running Android Apple obviously are concerned about privacy. So they're acting really, like, they're doing a really good job trying to protect us, if you think about it that way, at least from each other. They want to steal your data. But I think what's interesting is, I mean, the guys at OpenCloud are also now trying to roll out security fixes like on a regular basis. I think that's one of the things that he kept tweeting about. He kept saying that he's getting a lot of security researchers to look at the code base. That's one of the reasons why it's open source, I'm assuming, where they can patch stuff so that, you know, we can't be abused. That's the whole concept, I think.

Srijan Shetty: And the fact that most people can't install it is a very good thing because you're just preventing the hammer from falling on your own feet.

Utsav Somani: Somebody will solve it and build the solo $1 billion company that people keep talking about. But one final one. Can a bot actually, suppose you unplug, can a bot actually live and have consciousness even after that, after getting unplugged?

Srijan Shetty: So if, like, fundamentally, if you give it money, it can just clone itself another digital ocean box and just keep running.

Arun Kumar Gandlur: Boy. But what is consciousness? That's the question. What is consciousness?

Utsav Somani: Consciousness is that it lives without you having control on it. Like, I think for it to think freely and, like, live on the internet somewhere. I'm getting into the sci-fi world of movies, but that's yeah.

Srijan Shetty: There is a path to it. Like, if you give it a credit card, it can obviously just have its own subscription on some hosting provider, clone itself, and it can run forever. Like, it's not impossible. It's in the realm of possibilities, for sure.

Utsav Somani: All right, guys, we have to end this segment before it gets too dark. Thank you so much for coming on the show, man. Thank you, Arun.

Dhruv Sharma: Thank you so much. Great seeing you guys and thank you so much for sharing.

Utsav Somani: Oh, that got dark too fast. All right. Let's switch things up. Let's welcome our next guest, Aparna. Welcome to the show.

Utsav Somani: We're not exactly Shark Tank, but we will have fun. Thank you so much for joining us. Your episode was really nice. I saw Arjun Vaidya. I think he came to attend your viewing party as well. That's what Instagram told me. Tell us about the experience. Like, what does the company do? How is the Shark Tank experience?

Aparna Saxena - Founder & CEO, Antinorm: So, Antinom is not that old. We started or at least came to market in July of last year. So, about six months, seven months into the market. It's really a strange way. I mean, it's a strange way to approach cosmetic and beauty industry. Everybody thought at the time to build vertical companies, get into one specific domain, whether hair, skin or makeup, and just build it throughout. And I just came in with a very Antinom approach, now I can say, which is actually find the right audience, which is this fast, busy, working woman of India, and solve for them in every pain point that they find throughout their day. So, I made multifunctional products, almost like a SaaS approach, to be honest. And I think that's the tech, ex-tech we see in me talking, where I just kind of built a formula and then kept adding features to it. So, these multifunctional formulas are utility-led, so packaging that can be taken on the go, that can provide instant results. And I think that really resonated with everybody as soon as we came into the market. It was very differentiated, that POV towards beauty, almost like the quick commerce of beauty. And yeah, that was really interesting to see on Shark Tank. Now, the world gets to see it, where we come from and what we've kind of built. I'm really excited because we've seen massive traction pre and post Shark Tank. So, a good year for us.

Utsav Somani: And you've changed the name a couple of times as well for the business. Tell us, walk us through that journey.

Aparna Saxena - Founder & CEO, Antinorm: So, I think when it was a project, it was called Myriad, which was before I understood the utility angle, which came much later. Initially, it was just multifunctionality and time-saving. And then I picked a great name. So, true to my VC world, the name was Antithesis. Unfortunately, came across a lot of trademark issues with that particular name. And we landed up on Antinome and we haven't looked back since, I think it's the most fitting name yet.

Utsav Somani: So Dhruv, Aparna was working with Arjun of Good Capital. So, now we know the secret to Arjun's good skin when he was on the show as well. I'm guessing he's using the Antinome products.

Aparna Saxena - Founder & CEO, Antinorm: He is actually on my skincare routine for the last three years, which was what I built before Antinome. It was just a hobby. So, everybody at Good Capital was on my skincare routine.

Utsav Somani: What is that skincare routine? Like, I mean, for anybody who's looking to get good skin in a short amount of time, like use Antinome products?

Aparna Saxena - Founder & CEO, Antinorm: I mean, use Antinome products for just getting from point A to point B very quickly. They don't really fix any deeper issues like acne or pigmentation. That's not our promise. And I'm not a dermatologist, I want to say, but that skincare routine kind of developed in COVID. I had transitioned back from the US. The climate of India wasn't sitting well with me, so I developed it using what was available in the market because dermats weren't able to help me. And it was just a very simple blend of hyaluronic acid, niacinamide and azelaic acid in a set way every day for 30 days and it kind of fixes all the issues that you may face and may not be from what you're eating or hormonal. And it works really well for people who are just because of the climate struggling with this and actually that particular climate angle has carried forward to Antinome. So all our ingredients are for the Indian climate.

Dhruv Sharma: How do you define your consumer Ziparna?

Aparna Saxena - Founder & CEO, Antinorm: The way I defined it is actually not my consumer. So that's actually a good question. I want to tell you I went into the market thinking very Gen Z product, utility led, you can hook it on your bag, maybe it will have some vanity associated with it. But actually what my product's use case is super relevant for women over the age of 35. Mothers, homemakers who don't have time, working women, travelers, and that's actually been our demographic. So my largest contributing demographic is women over the ages of 35 to 45 and 45 to 55 and then followed by Gen Z. And this particular woman has little to no time in her day if she wants to really maximize it, if she wants to stay active, go to work, be a part of her children's life, do networking and still have some time left, Antinome comes in and helps her through that.

Dhruv Sharma: I was just saying you're merely simplifying life for them. And I was sharing a tweet by PG, Paul Graham of all people on this with someone on our team, right before we started the stream today where he said what's his wife's name? Jessica, yes. Jessica and I in our restroom maybe they have a twin sink set up. He's like my side just has maybe a toothbrush, maybe a razor, something. And Jessica's side it looks like a portion cabinet. And so sometimes people are like spoiled for choice. And again, this all the stock around the three, five, seven, 10, 15 step skin care routine. Nobody has the time for it. And you're trying to simplify that or maybe just tell us you use two words frequently which is utility and multifunctional. Can you help us understand what they mean in practice and effect?

Utsav Somani: Yeah. So maybe I can run us through two or three hero products of yours and how much of the time saved. Like I think I'll sort of quantify it for us.

Aparna Saxena - Founder & CEO, Antinorm: Yeah, no, absolutely. And I'll address that as well. We did a survey of women over the ages of 30 and nearly all of them came back and said they've bought 10 products, but they haven't been able to sustain that routine more than one month. So what in fact happens is women are marketed these products and they kind of grab a hold of them, but they're not that's why retention is bad in the beauty segment because in fact women are not able to sustain these hundred types of products. And my thesis was you know how we see companies very often like this is not a company this is a feature and we saw a lot of those companies come by during COVID. I felt that the cosmetic industry was inundated with features and that was keeping that in mind came my first product which was it's just a hair serum but we've added ingredients for the climate like avocado oil which prevents oxidation high moisturization which makes frizz go away. We added other ingredients that make it a heat protectant and then we kept on adding it just like I said like we found the base model and kept adding these features like a SAS model and it replaces about 11 products with this just one product. How does this save time? If somebody takes a shower and applies it on the hair and comb it and go straight to work on your way of your commute your hair will dry representatively versus earlier women were spending about 30-35 minutes blow drying their hair damaging it and then in that vicious circle where they were continuously buying more products to fix it so that became our first hero product. The second one which is my favorite it's called facial in a flash again base model of something that does very for sensitive skin does very small amount of scrubbing on your face it's a rice enzyme peel that's actually a Japanese invention we took that base model and we added all the benefits of my skincare routine which is niacinamide kojic acid and it acts like a full facial so one hour at the salon or one minute with that facial in a flash on your skin removes any dirt dust pollution for the climate of India in a packaging that is so utility focused instead of those tubes and other things you just press it it kind of comes on your hand so you don't have to like really work very hard doesn't leak in your bag fits in your cabin luggage all of these literally every part of that consumer journey broken down so these became our hero products at 30 minutes one hour and similarly with other products that have come henceforth

Utsav Somani: but I mean keep adding these ingredients do the price does the price increase accordingly

Aparna Saxena - Founder & CEO, Antinorm: no actually yeah I mean it does increase and that's why we're in the price segment of about I think our AOV is high but our average selling price about 1100 and that is why we're not able to sell this at say a 500 today because these one innovative packaging and then mixed intermingled with these really high quality ingredients whereas other competitors just sell one ingredient literally in a bottle can price it at 400 so yeah it does increase the price but we've been able to kind of create blends that we still remain a good high margin company at this point in time only to get better

Dhruv Sharma: how many people are you sorry

Utsav Somani: there was a clip about your formulations about AI being used for it I saw that on your Instagram just before the show is I mean Aman was I think talking to you about it that how could you use AI to do these formulations what's the story behind that

Aparna Saxena - Founder & CEO, Antinorm: it's really funny it's the catch so we when we started building Antino we weren't a cosmetic company we were a data modeling company we have two data models one's a trend based data model where we look at 10 variables from around the globe to understand where the market is headed and the second is an ingredient based data model and it actually helps us understand what the Indian market ingredients are and we've kind of built this data model taken it over to the AI trained it to be able to understand what is available near us what is optimized for the Indian climate and then we give it to a formulator so it's not really formulated on AI but it's you know what I look at it as the 180 days that I would probably spend going around the market and talking to people and figuring out what works it's reduced that headache for us and it still goes through exactly it still goes through a formulator that goes through microbial testing everything and gets to market the same way it just feeds us along so from inception it was 30 days because of this very reason

Dhruv Sharma: sorry I interrupted you no but this was going to be my question I was going to ask you how many people are you at now and it seems to me there's a couple of people and then a whole bunch of AI

Utsav Somani: but no you answer that question for now those are getting scary

Dhruv Sharma: don't worry about them

Aparna Saxena - Founder & CEO, Antinorm: so we were only five till last month we've now graduated to 10 of us yes

Dhruv Sharma: that is so congratulations on that and you know of the five products that you mentioned maybe just pick one look I am completely oblivious to everything that you just said right like who are formulators what's the process from start to finish of bringing one of these beautiful products to the market

Utsav Somani: appreciating that I think she's gonna order after the show ends

Aparna Saxena - Founder & CEO, Antinorm: I love that that's so great in here so our trend based data model runs in the background we're we're obviously not that far like I don't think it's perfect yet yes some of it has worked I was actually entirely conceptualized on that but right now for example the company has about 10 to 20 products that the data model has given an output for gone through the ingredient base and then with formulators we are on our end are trying to figure out and talking to customers understanding what is the biggest pain point and that's what we'll get to market but I think from a time that we're able to get a positive understanding from the customer to getting to market it's we can reduce that timeline to as little as 30 days which is actually our core strength speed we're able to kind of really really sweep things along and almost always or at least I would hope that we get there be the first to market with anything truly innovative and because our packaging structures unlike other beauty companies where they want to put it on a nice pedestal and they all look like each other and it's very quiet luxury we decided you know what it doesn't matter if it doesn't look like each other as long as it is actually solving a pain point with the packaging as well we'll do that and that's really somehow worked for us

Utsav Somani: and given the competitiveness of this market I mean I'm sure you must answer this question a lot that how long does a competitive advantage last for a company building in a direct-toconsumer say beauty space what's that answer for you

Aparna Saxena - Founder & CEO, Antinorm: today it's about building a brand it's really not about anything else and then because we are a D2C company we're not bound to anything today if one of I launch a product and say it doesn't work it's it's not getting the love that it needs for me to discontinue it and focus all that energy somewhere else super easy to kind of redirect my energies there and then keep on building till we find those five heroes that actually make us a big enough company and I think the data part is what's going to help us get there and the speed of execution is what's going to get that so that's become the real moat in my view

Utsav Somani: and I mean team is always digging out interesting tidbits so you built an all Gen Z team and wrote about it it's like going from type B charger to a lightning bolt how is it working with Gen Z's

Aparna Saxena - Founder & CEO, Antinorm: it's super interesting they work very differently from us millennials I must say I think it's funny

Utsav Somani: that you assume what if we're millennials what if we're like Gen Z just looking older

Dhruv Sharma: or boomers looking younger

Utsav Somani: that's too much dystopia reality for one show yeah

Aparna Saxena - Founder & CEO, Antinorm: I mean they're super interesting to work with they're all like they really focus on the softer side of things

Utsav Somani: we were really busy unfiltered show by the way so you can say truth

Aparna Saxena - Founder & CEO, Antinorm: I mean there's definitely days of not feeling up to it and has happened from people today today I'm not in the mood to kind of work but I truly think that they are more intent led that's the true answer they're more intent led we're more focused on getting things done and you know just doing them the proper way they're more intent led they really want to make a difference and that's kind of helped us along the way even when we had the name change so we were launching in two days and our name changed and suddenly everything went away flying out the window they're super resilient because they're super plugged in they're super hopeful versus us we're like okay great another dead end let's start over so it's been it's been interesting but very different very different from working with millennials

Utsav Somani: so any final closing question for Aparna

Dhruv Sharma: oh well I mean maybe just the broadest one which is what's coming next Aparna or also how's for any other young founder who's looking at you right now while you're still very early in your journey and going on to become like a role model how is you spending your time and how's that changing from step to step stage to stage yeah

Aparna Saxena - Founder & CEO, Antinorm: I'll show you a very cool product actually this is for shower in the morning after the gym and go straight to work it hooks on to your back so this is really cool product it's really picked up and we have some more interesting stuff coming in the pipeline that's coming in the summer so that's what's next I can't say concretely we're six months old so I can't say concretely

Aparna Saxena - Founder & CEO, Antinorm: happening but as a founder I've realized the biggest pain point of my life and existence is which I did not anticipate at all so I'm spending a lot of time on content and talking to like like speaking on social media and that's been a very big change from my desk job to this and it doesn't feel like work someday so I feel like I'm not productive but that's changing now and I mean increasingly as a founder I think prioritization of work is really something I struggle with and that's the candid answer so anybody coming into it don't expect for structure which is what I did that was my big mistake I expected for a lot of structure I'm setting something up so I bring structure to it that's really if you're building in this fast-paced competitive phase that's something you're not going to get no matter what.

Utsav Somani: Awesome. I have one more so friends there was an article mentioning the use of peptides in the US and I mean these grey market peptide market that have popped up in the US where people are experimenting with different kinds of peptides have you looked at that?

Aparna Saxena - Founder & CEO, Antinorm: We've also looked at PDRN. We consistently look at like what could be an interesting ingredient for us and if you know we can use it in our products. So far we haven't seen any use case for it in our products because we're not able to fix you know skin barrier and things like that but I do we are working on something interesting where maybe that we could find a blend of it mix a skincare routine and bring it into one. I have positive I think peptides are great by the way they're fantastic and they should do really well. I think that's going to be the next big trend skin barrier and peptides for the next this year specifically.

Utsav Somani: Awesome. Wishing you all the best and congrats on the new milestones. Yeah.

Aparna Saxena - Founder & CEO, Antinorm: Thanks for having me. Thanks so much.

Dhruv Sharma: Congrats. All the best.

Aparna Saxena - Founder & CEO, Antinorm: Thanks so much. Bye.

Utsav Somani: All right listeners that's it from our guests. We have a couple of news items to run past you. There was the union budget which was announced yesterday. So did you get a chance to review that a little bit here and there.

Dhruv Sharma: What are the big headlines itself?

Utsav Somani: The big headlines. I mean not much on the personal tax front. They've done a few things. They've reduced the TCS a couple of things but I think the major headline that we cover a bunch of topics India announces zero tax until 2047 for cloud companies setting up data centers in here in India and using Indian resellers. So that was a big one. I think that clearly there was a big push by the Indian government to bring these large data companies and set up cloud centers cloud spaces here as well.

Dhruv Sharma: And many experts are reading that as the government's you know signal to the market that India is going to be like a nominal place for big tech companies to come and build data centers serve even global customers from here. And I believe that tax already extends to global revenue. They would earn a lot of money out of supporting AI workloads here in India.

Utsav Somani: And then of course Google and yes. And stuff. I think it's pretty big.

Dhruv Sharma: Yeah. It's big for them.

Utsav Somani: Yeah. And then there's I mean a big push to sort of make young Indians spend time more productively. There's syntax and cigarettes and exercise hike is going up. Liquor exercise keeps going up. There's a complete ban on real money gaming. So a lot of these activities have happened. But they're also giving the same treatment to FMO trading where the STD the tax on these short-term trades have increased 2.05 percent. Options have gone up 2.15. So I think they're trying to like sort of curve speculative activity on the stock markets if that makes sense.

Dhruv Sharma: It does. This is something we covered I think in one of our earliest shows as well right when they announced a ban on real money gaming right now there's the explicitly stating that we want to put an end to speculative activity and redirect all of that energy into more productive activities. So yeah, there's a common thread running through all of this. The brokers aren't happy. Mumbai is not happy. The trade is not happy. But you know, I mean if your government you've got to do what you've got to do.

Utsav Somani: Long term I think it's good. Maybe let's see. But buybacks also now taxed as capital gains. Good for promoters. Good for shareholders.

Dhruv Sharma: Yeah, buybacks have gone through a journey of the last two or three years. Like I suppose three years ago it used to be that the company would pay tax and whatever it was going to you know then transfer to you and you would receive those considerations sort of tax-free. And then they were treated as dividend income for I think one financial year and they're going to be treated as capital gains now. However, the cost of acquisition of the asset will be treated as a capital loss so you can use it to set off gains or even carry it forward in subsequent years. And for non-promoter shareholders the rates is just the standard capital gains rates. For promoters and I mean there's an additional component. And the effective tax rate I think goes as high as 45% if it's a foreign company.

Utsav Somani: So that's happening at pretty rapid pace. Like I think sovereign gold bonds which I think were quite popular with investors as well were supposed to be tax-free at maturity if even bought from primary issuance or secondary markets. But suddenly like now the government has said that they'll be taxed if you bought them from the secondary market. So a lot of changes keep on happening. I think it's good to keep a tab on all of this but get a good tax advisor guys. Yeah.

Dhruv Sharma: Absolutely.

Utsav Somani: All right. So we're supposed to do a newsletter plugin. We're launching a new short newsletter which will be with our community from starting tomorrow. So check out our social media and WhatsApp community which are there on our website for more details. And we'll see you on Wednesday. Thank you for tuning in. Bye-bye.