E56: Building AI from Bangalore with Yash Chavan

I was like looking at my schedule and sleep patterns for this week and sometimes being a CEO or being a founder is hard. Sometimes. You think so? Every now and again, I'm like, like I woke up this morning at, this is the, I think this is the one of the few times where I like really felt it. I had to go to Cincinnati for a conference yesterday. And so I, I like, I like touched Denver for 20 minutes on the way there. So I like got outta one plane and then got onto the other. I was like on Denver Airport for literally 20 minutes from plane to plane. Ended up with 3 20-hour days. This week. 20-hour days are not easy for what it's worth. Like I work pretty hard, but a 20-hour day is a little bit harder than what I usually do. What's the typical day like? Like 12, 14, or? Yeah, yeah, yeah. It's closer to like 12 to 14, which is like, that's like a normal fucking cakewalk. Like, are you including like, like kid time and stuff like that in there? Or just like pure work hours worked? So I wake up at 5:00, um, and then I'm usually doing stuff like while getting re— I mean, AI has changed how much you can do parallel processing, so I'm usually doing stuff till like 6:30. Then I help my kids get ready, drop them off. That's usually done by 8:00. Um, and then I'm like straight working 8 to 5, uh, or more like 6, and then with the kids till 8 and then working again till like 9:30. And so the average day basically is like 5 to 9:30 with 3.5 hours taken out for kids stuff. So if you like roughly do the math on that, you end up with between 12 to 14, right? Because like 5 to 5 is 12. Add another 4.5 hours, subtract 3.5 for like hit time, right? So that's like roughly how the math works. And so, uh, but yeah, this week I, because of travel, I had like a few 20-hour days and those are hard. Those are hard. I can imagine. Is this, is this 4 hours of sleep a day? You're just like into work and then, well, yeah. So on 2 of the days I had 4 hours of sleep. Uh, so yesterday was one of them. So yesterday I had 4 hours of sleep after a 20-hour day. And yeah, the intensity is also high. This is the other thing. So when I was in Cincinnati, I think I was like talking to people nonstop, which, I mean, the conference was super productive, but it's like not low intensity. Do you know what I mean? It's not like I was there and hanging out and listening to— I didn't go to a single session at this conference. I was like talking to people the whole time. In fact, I don't even know what the sessions were. So, you know what I mean? Like, and so, and so I like, I got home last, I reached home last night at 12:30, uh, and then I was like, and I like finally went to bed at 1:00 and then because I was like catching up on emails and stuff and then I, I woke up at a little before 6:00. So anyway, so I was like, man, this is, this week really, and then Right after this, I have 2 sales calls. Oh wait, you're in the sales call too? What's your drug of choice? What? I was saying, what's your drug of choice? Are you just doing this undrugged or is it coffee or Zinz or coffee? Okay. No. Well, I only do coffee and Diet Coke. I'm a wuss, man. I can't like anything past coffee and Diet Coke. I'm like, I'm too much of a wuss to do it. Unfortunately. Prioritizing life, staying alive, and, uh, not being addicted to things. I think it's a good idea. No, sometimes I wish some of my Wall Street friends, you know what I mean? I did a 20-hour day on— well, I'm on vacation this week, um, but there was this thing that TikTok wanted, uh, us to do, so I did a day trip to New York on Tuesday. Um, so it was 20 hours, but it was like I, I intentionally said I'm just gonna relax, so I listened to a book and just like slept a little bit on the plane and then it was literally like 12 hours, not quite 12. Yeah, close to that. I was in like in Manhattan for like 8 hours. Um, but it was intentionally a relaxed experience, so it's not the same as, as what you're experiencing. But, um, but yeah, I was one of those that like woke up at 4 in the morning, was in Manhattan by noon local time., and then, yeah, but, uh, it was, it was supposed to be relaxing and it actually was relatively relaxing. That's amazing. That's amazing. I'm realizing that I, I have a newborn baby, but, uh, I have a more relaxed schedule than you guys and I'm getting more sleep. I think that's an interesting thing. I mean, you know what, the one thing I would say from that I'm learning with having a newborn baby is context switching is pretty hard. You know, when you go from like holding the baby and relaxing and then to like back into it, there's a little bit of friction. It's like, you know, I'm used to uninterrupted, like, you know, I'm in the screen from morning till night. But, but yeah, there's a being a dad thing. It breaks up the day a little bit. Yeah, it's not, it's not your time anymore. I've also had sick kids this week. So it's actually been kind of nice being on vacation because normally that would just be extremely disruptive. But yeah, you know, when you have a kid vomit all over the place at 5 in the morning, it's like you start your day and then you can't go back to sleep really. You're just like, everything is just chaos at that point. Yeah. Yes. What is the int— by the way, one of my biggest learnings being in Cincinnati and Jeremiah, I'm curious your take on this. I'm going to say something that's going to sound controversial, but there was this tweet or X post. Where this guy, Elad Gil, he's like a famous investor, said like, the labs are like a month ahead of SF. That's 3 months ahead of New York. That's like 6 months ahead of the rest of the country. So something like that. By the way, like I have been saying this to my team for like months at this point and trying to get them to spend some time in SF as a consequence, because I'm like, I'm not sure how to, how to explain this to you, but when you're in New York, you just are not at the same speed, or like you're just not at the same point as like the majority of SF. Now, I think that tweet got interpreted poorly because people don't understand like all these things are normal distributions. There's always going to be people on the edges and all this other stuff. And so like it's a comment on the displacement of the distribution, obviously. So anyway, let's just assume we can all be pragmatic about that, um, and, and, and talk about it in that way. The— so anyway, I was in Cincinnati and I was like, dude, I, I think the rest of— I think the difference between where SBF thinks the world is and where someone in Cincinnati thinks the world is, is like 2 years apart, not 9 months apart. I think they're a year into ChatGPT, not— and I'm not saying it to be like negative about that. I'm just saying that the adoption speed is not that fast. I'm surprised at, I've been to a couple local things in Denver lately, tech things, commerce things, stuff like that. There's like a, I don't know, a Venn diagram, I guess, here of like commerce, like D2C, leading influencer types and tech leading influencer types that are overlapping. And you're seeing that, that middle layer is what I see in my day-to-day. And in terms of like, you know, my engagement on, on X and, and the people that I talk to, and that's like the, it's like the 1% of, or maybe not 1%, the top 5% of each industry is merging. And you've got like this little bubble basically that's way ahead of everybody else. And then I talk to other people who are in software technology, commerce, um, that are kind of behind that. And they're, they're just starting to use some of these tools and not nearly at the level— I mean, I mean, they're at the level of many of them, um, just now starting to use things, things like ChatGPT. And, you know, I'm going to rewrite the copy on my website with, with AI. It's like, okay, well, that's— yeah, that's like 2 years ago. Um, but I completely agree with you, Arsham. Like there, there's a, just a massive, uh, there we're so like, we're especially in the circles that you run in are so far ahead of the, the, the bell curve. Um, but I actually think that's a good thing ultimately. I, I actually think if everything was moving as fast as the, the early adopters on the front end are moving, it would be a negative for society, um, to move at that speed with these kinds of tools. Um, just because there's so much potential for disruption. And so actually the fact that the curve is longer than that is actually a good thing. And there's a lot of things that can be learned by the early adopters that can then be applied, um, later on in the curve. Hmm. Slower speed of change. Yeah. Uh, what do you guys think about this? I mean, one of the problems I had with my first brand was that because they were downloadable products, uh, like after we had a few hundred sales, you know, we'd be lucky if we got to like 1,000 before this happened. All these like Russian piracy websites would have my product freely downloadable. And then our competitors would use those and basically resell them. They would change them slightly and, you know, make it look a little different and then resell it. And I remember thinking like, I did all the hard work. I didn't, we were kind of a copycat, but I remember thinking, man, you're on the shoulders of all this work that I put into this and you just get to like change it up a little bit, take it to market and you've got yourself a brand new product. And I kind of see the same thing happening with the LLMs. You know, like Claude makes an advancement, accidentally leaks their entire thing online, and everyone is caught up to Claude very quickly. I just, I wonder how much advantage there is. And then if we go to the micro, which is adoption of these tools, I wonder how much advantage there actually is in being a first mover here, if the time to catching up is always so much faster, right? Like, so if we zoom out and we say, okay, typical software business, SaaS, right? Your moat was the actual ability to build. You needed developers, you needed people who could actually build the thing. And then, you know, obviously that moat today is not so much of a moat. It's not a moat at all. And people can build something that might have taken you 6 months, a year, it might have been completely defensible, like maybe in an afternoon, right? This like time to catching up is so fast. I remember I started building my app in like September, you know, August, September last year, right? In Cursor, and it was painful, right? And now Claude is so good. Codex is so good, right? That like, it would be much, much quicker. It wouldn't be quite so, uh, much of a slog, right, as it was then. And even then it was incredible, like how much you could do, right? But today, you know, it's very different. So I wonder, yeah, what are you guys' thoughts on this? Like, I guess the first mover advantage and the time to catching up really just closing down and eventually, you know, people being able to catch up very, very quickly, right? And do some incredible things with these tools despite not doing it since, you know, Q4 last year. My perspective on this is you can only answer that question if you know what you're trying to do with the business. So let's just take a venture-funded business, um, like ours. Then speed and being at the very beginning matters because the opportunity cost of not capturing the market share and making, trying to make the switch cost higher and higher. So let's just take Harvey and Lagora. Because it's like the easiest way to think about this exact question. Because what they did is basically spend immense amounts of capital to quicken the speed of adoption of legal at a rate that is unnatural to the legal world. And the way they did that is they spend money to hire lawyers and then just like literally deploy and have like huge CAC in the deployment on the first year to like acquire every customer, but there's like, whatever, some like finite number of large firms. And so you want to just go and land grab as quickly as possible. And then it's always harder to switch later. Um, like no matter how shallow it is, it's still annoying to like switch your workflows. Um, and so, and the reason is those businesses were set up to, uh, like the strategy was, hey, let's go and transform this industry, the legal industry, by giving them agentic tools to run all of their workflows. Uh, and then you have to go super fast right now. A different way of asking your question, or a different strategy, Jack, I, and I think that's sort of what you were saying is like, say you're running a business whose job it is to cash flow, right? Then, then the way you would want to do it is you want to want to optimize for how much cash flow you're generating. And then actually maybe it doesn't matter depending on how you're playing that game for that period of time. But I think that the, the actual thing here is you got to be really clear-headed about what your business strategy is and Where it basically nets out is if you have shareholders who are not the operators of the business, what it means is you have a duty to improve the value of your stock over time versus generate cash flow. And the consequence of that is, which is like, again, a strategy thing. You can decide not to do that. Right. Uh, but as long as you have that and you have taken on the responsibility of improving enterprise value over time or improving stock price over time, you have this terminal value question. And then what that leads to is a question of like the land grab that's happening right now. Um, and so like, I, I think it has to work backward from how you're, how you're set up. Uh, I also think like the industry and the geography matter a lot. And I've, I've been curious to get Yash's view in for a while on like the speed at which this is happening in India. Cause like I can tell you that in the U.S., I mean, first of all, in SF, the speed is just insane. Like you would think that AI happened a year ago, right? Versus is happening just because literally every billboard is AI, right? And so it's like, oh, of course this is like already a consensus thing. Um, but I'm curious, like, yeah, in India, how it's playing out and then Yeah, how do you think about how much it should impact your operation, how much it's impacting the speed with which you do things? I think, yeah, we're in a unique place because we are operating out of India, but then like are sort of also like operating in the US. Like we're up against somebody, like some one of our biggest competitors actually is based in San Francisco, right? So we're in an interesting spot. I think like geography-wise, India in general nowhere near SF. I think I'll probably put it somewhere near, but like not super lagging either. Like there's tons of great AI companies and models. I don't know, Rishabh, you've probably heard of Servam, right? Servam.ai. Yeah, like, so there's like a lot of these modern companies in Bangalore, a lot of new like application layer AI companies that are, that are like on the cutting edge. Some of my friends are founders of some of these companies. So yeah, I think like there's a lot of movement happening here as well. But yeah, nowhere near the— nowhere near SF level. I'm sure like, again, it's like normal distribution. There's probably some people at the edges who are doing things like that. But then yeah, I think like the normal average also like, I'm also strictly talking about like tech people, right? If you look at like the average person, like in working in consulting or something, they're still probably like, they're using ChatGPT to write emails or like negotiate a raise with their boss or something like very basic 2 years ago use case of like just chatting with the thing. But then in tech, I think it's definitely like it's picked up, especially like in the last few quarters, I would say. I think that's underappreciated and why people got tweaked on X when Elad made that comment. People forget that the average person in SF is a tech person. Yeah, there's no distinction between an average person in SF and like I walk, if I walk on the street right now outside and I turn left and I say, hey man, how did you set up that Codex Agenda workflow? And I don't know the person, they would respond to this with an answer. Awesome. As I mean, as long as they're not using the street as a toilet, right? That's just luckily not yet the average person. We're not yet at that point. I think that's like two different sets of people, I would say, right? Yeah. Yeah, but like a lot of these companies, I think Anthropic now has an office here in Bangalore. Cursor has an office in Bangalore now. So I think a lot of these like model companies are also opening up, which is, which is a signal, right, to interpret. So yeah. Finally, maybe we're going to see the truly global workforce, right? Well, I mean, I don't know if that's a, that's a good thing for the US workforce, or, you know, especially in tech. But hey, you know, it's a good thing for hardworking people, you know, location agnostic, right? Yeah, but I think a lot of the whole transformation with of India also has been just like there is a push to go from services to software. I wouldn't say like a push, but I think like a natural step in that direction because a lot of the whole India tech from the '90s, early 2000s was just like services ecosystem, right? Infosys, TCS, like all of these companies. And essentially it was just like cheap labor, right? It's like the most crude way to put that. But then now I think it's moving more towards like software more so than— I mean, services is still a large chunk of the economy. I think like a large chunk with the whole like our startup revolution really like started maybe in like early 2010s. So maybe like 15 years ago from now, or some of like the biggest companies in India who are now like public companies started and were startups. I think like since then a lot of it has moved from like people who would typically be at one of these services companies now work at a startup, right? Like the makeup of that person. So like the fresh college grad, like when they would like 15 years ago go join Infosys and that was like their dream job. Now they want to work at a startup in Bangalore or in Mumbai, right? So like I think a lot of that is changing and moving. Incredible. Amazing. Yeah. And the sentiment, like excitement, like from the people that you speak to and not just like obviously your friends that are founders, but like the people who are, you know, getting started, excited to speak to you, excited to hear your thoughts. Are they excited or is it like— Yeah, there's so much optimism. Yeah, I think like everyone's as bullish as we've ever been. So yeah, about the space, about the ecosystem, about the country, I think you name it. I mean, Obviously, like, there's problems, but people are way more optimistic than even like 5 years ago, I would say. So yeah, especially with AI now, because a lot of the things like, if you can, because a lot of the delta used to be how good is your code, but then if now the agents are just coding, now it's really about like strategy and all the other things. So I think like it's a great equalizer in some ways. How do you think about growth rate and like the product launch that you just did and launch schedule and, and things like that. Like one of the biggest shifts to tech marketing in the US is product velocity is the number one reason why, at least in the enterprise, why a company works with another company. It's not even actual, like number two is so far behind that it's almost I don't even know what it is anymore, actually. And I think it's because we are in a shifting landscape. And so when you're in a shifting landscape, your buyer has to have confidence in your speed. Yeah. Otherwise it's like, I'm gonna buy your tool and I'm gonna wake up in 6 months and I have to go buy something else because like the thing has just changed. Like the, what I need has changed and what's possible has changed. How do you think about it for you? And yeah, what is the general— and amongst your friends in India, like, I know here we— that's all we think about is shipping velocity. Yeah, I won't say that's all we think about. Also, like, I think a lot of, like, the ecosystem that I'm in still is not enterprisey. It's still like mid-market SMB, and they still care about velocity, but maybe not as much. So we do get asked questions about how quickly are you shipping, what was— what did you ship last month, what does— we sometimes have to like show our changelog on sales calls saying like, okay, like here's all the things that we shipped last month, here's what's on the roadmap, here's what's coming, blah blah blah. But then definitely not like— I won't call it number one. I think it's definitely top 3, top 5 though, depending on the prospect that we're speaking with. And that's that's definitely changed from even like a year ago, people wouldn't really care about that. But now they care about like, hey, do you have an MCP? Do you connect to like, does your app connect to cloud, things like that, which is also part of the reason why we shifted and launched like a new product. Because the old product was just like legacy code written by an agency, endless tech debt. I mean, to turn that into an MCP would have been like a lot of work, grunt work. So just like having a new clean software product that's just like built in the right way, that's almost almost AI-first in some sense, also made sense from that perspective, which is why we launched, did that whole launch and then spun that out. And there's like other business reasons for that, but then that's definitely one of them. Jeremiah, what's your sense on the importance of like shipping velocity right now? Yeah, it's interesting because I think the, it kind of goes back to what you were talking about with earlier with what is the business objective and goals. So I think if you're talking about like a VC-backed company, it's all about momentum. Momentum matters so much because you're trying to get to the next step effectively. You've got like this series of steps that you're trying to get to. Whereas, you know, in the context that we operate in a holding company, it's really just about like sustained profitability and profitable growth over time. Growing EBITDA matters more than anything else. And so, That's a— so anyway, all that to say, I think like the, the, you have to understand the goals of the business and what the shareholders want. And then in that context, then you have a better understanding of, of what to do. So I think, I do think shipping velocity is really important, but I think there's a couple of things with that too. One, pitching costs is still a real thing. So you don't necessarily have to, to be the absolute leader in features. Somebody else could ship faster, but there's still like a cost of getting them to switch off of whatever the, the tool is that they're using. Um, I mean, we experienced that early on actually. We were way ahead of the incumbents. We kind of came in a little bit later than a couple businesses. We were way ahead of the incumbents for like a year. Um, but it actually was really hard to get people to switch off of those tools. That, that took time. That was a process. And so now we've kind of I would say we're, you know, probably the first company that most people think of in this space on the, the no side for surveys. But that took years to get there, even though product-wise we were actually there long before the reputation caught up. And then I think the other thing with software that's really important is I think this is going to become more important over the next couple of years is also the stability aspect of it, however you want to phrase that. But basically How secure is your product? How well does it work? Can somebody actually trust that it's going to deliver on, on what's promised? And I, and the reason why I think that's going to become more relevant is that there's just, we're starting to see more hacking happening. We're seeing like, you know, there, it is really easy to launch something new. You have to be very careful about the, the underlying performance, stability, all those kinds of things and, and taking a piece of software and scaling it rapidly. You end up running into issues that you weren't expecting, uh, to, to pop up. So I think there's a balance there, I guess is what I'm trying to say. And ultimately then it comes down to what, what risk profile do you want to have in your business? What are your objectives in your business? And then shipping speed, I think depends a little bit on, on how you're thinking about those things. How much, how much are you willing to invest in shipping faster? All those kinds of things. I'm so, to be fair, I'm not either. I'm actually very curious to see how this ships. Turns out. The, in the past, that was a big thing, but there's kind of a shift that's happening in expectations. I mean, Claude is down almost every day, it feels like. Um, yeah, there's like, there's all these things where, uh, I do wonder if there's a certain rate of shipping at which people don't care about stability. I just wonder if we get tired of that after a certain point in time. Like if, if you have a piece of software that you feel like matters and that you need to rely on, do you get to a certain point where 6 months of having this thing, uh, be out for, you know, half a day every week, do you sit down and say, I just can't take this anymore. I just need something that works. Um, so that's, that's where I'm coming from there. I think that that may be everything is kind of a pendulum and it swings, right? We've swung really far toward. You don't think so? You don't think there are software where— so, so a lot of software, um, iCloud, for example. Cloudflare. Cloudflare can't go down. Okay. Cause Cloudflare, the tool availability. A good, so I'll just use surveys. So when we are, we are showing a survey on an order confirmation screen after somebody places a purchase. That is a singular moment in time that you have the ability to actually have that connection with the consumer. And so, and there's, there's some exceptions, there's some percentage of people that come back to a thank you page in the future or whatever, but largely speaking, you have one chance. And so we care a lot about making sure that we do not have downtime in that experience. And of course it happens a little bit. Shopify goes offline, right? But we actually have better availability of our post-purchase surveys than Shopify checkout does. And that's kind of a thing that we, we look at and I think matters. So you could go vibe code your own piece of software and if it stops working, you might not know for a week. And so in that time period, you missed all of that data, right? And for us being the builders of this thing, if we go and just ship things without any— I shouldn't say without any consideration, but there's a speed of shipping in which we are risking that experience breaking. And if we break it too many times, I believe our customers would lose trust in our ability to actually do the core thing that we say that we're doing. I don't know if that makes sense, but that's— I think there's like certain— not every software maybe has that problem, but I think a lot of pieces of software, the whole point of software is that something is happening automatically for you that you don't have to consider. And so if it's, I need this thing to do my workflow and I'm participating in that, I think that's a different story than I need this thing to just work. And in that context, I need availability to matter, um, as well. So yeah, I think the simple question is, is tool availability directly related to output? But the answer is not yes for all agentic tools. And that's exactly the insight. The, the reason why it doesn't matter for Claude is it kind of doesn't matter. Like the availability of cloud does not actually have a direct one-to-one relationship with your output of you using cloud. No, it's true. Yeah. And I guess what I was saying, the, I think the stability, availability, whatever does matter in software and shipping speed and those things are inversely correlated. I mean, there's a, there's a, not always, but, but not for Agentic Workflow software., right? No. Yeah, yeah. I, I agree with that. Yeah, yeah, yeah. Oh, I wish I knew what you guys were talking about. I don't know. Claude went down yesterday and I was like, fuck, Codex is kind of performing better than Claude recently and now this happens. So I switched. I was like, okay, everything. So when I start a new project to start a new thing, a new terminal tab, it launches Claude in my project folder. Now it doesn't, it launches Codex cuz Codex has been outperforming for a little bit. So I'm like, okay. And then Claude went down and I'm like, what do I do? How do I, I can't use Claude right now. So don't get me wrong, I haven't canceled my Claude subscription. I'm still paying for Claude. I'll probably switch back to Claude if they pull ahead again. But like right now, everything's Codex, like from default. And if they, if they were offline for enough time, I would cancel my Claude subscription. Yeah. Well, I think once you start to get into like the bigger enterprise stuff down the road, like let's say 5 years from now, I think that stability reliability will matter more for those tools than it does today. Because you're going to get into that world of like, I'm picking this tool, I'm building all of my infrastructure around this tool, and if that stops working, our business stops working. But I, I should say that I'm like, I'm one, one person, you know, Claude goes down for 5 minutes, not a disruption. Yeah. Think about human availability. I'm like so poorly available. I don't answer my Slacks. I don't, I'm only 20 hours a day. I'm fucking tired. I'm like, and you guys are talking about software availability like it matters. We pay like so much more money for humans who are poorly available than software. But, but isn't that point part of the point of software, right? Like if your Shopify checkout goes down for 4 hours and you're a brand that does, uh, $600,000 a day, that's $100,000 you just lost because your software went down for 4. Yeah. I'm not disagreeing with you on, on that type of stuff. I'm, I'm simply remind, like, I don't wanna lose sight of this idea that there is now a new form of software, which is agentic workflows that does not have the same property as the previous generation of software. Where I can give it a task and it can do it when it does it, and it'll report back to me when it reports back to me, just like a human does. And in the middle, if it was down for a couple of hours, it makes no difference to my life. Yeah, I hear you. Yeah, that makes sense. It's the same thing with the token cost thing. It's like, well, this is infinitely cheaper than a developer or XY, any individual person. I was like, well, I don't know how much I care about token cost going down. I, I really don't. At this point, but I, again, I don't have 100 people on my team using it. It's just me with like 2 $200 subscriptions. So I'm like, yeah, fuck it. Use as many tokens as you can. Wait, guys, actually, can I take this opportunity to shift the convo to this new thing that I want to talk about? Because I've like fucking gone crazy with it. Yeah. Have any of you guys used the Codex, like cloud-hosted agents? Like that you can like create your own agents and it has a, it like, it's a very nice UI. So like it'll automatically set up a memory file for you. It'll like, it shows you where all the connectors are. And so like, it's an, it's a real agent workflow. So it will like improve as you give it feedback, basically. Okay. So anyways, talking about availability, which is why I fucking love this. My team, I've become, what is the number one problem with a CEO? You can't reach them when you need something. Exactly. Okay. So I made a Rishi bot that responds to my team's questions the way that I would. So I gave it a bunch of training material, which is like my previous ways I respond, things like that, blah, blah, blah. Because I have big downtime. You think Claude has downtime? I'm just like not available 99% of the time. Right. Um, and so I create this bot and I tell it to use its memory file such that when I respond to it and I invoke it and I tell it, here's how I would've responded. It continues to learn. So every interaction it has, it gets better, right? Because I'm giving it, I'm giving it constant feedback, but it is learning better than a human because it just directly takes it and puts it in the memory file. Like, I don't have to remind multiple times, right? It puts it in the memory file. Now it's in the memory file and that's it. Dude, this thing is astonishing. Astonishing. You guys should build one of these bots. Go to the Codex, like whatever, whatever they call these, like cloud-hosted agents thing, and create a bot that is basically a version of me. I created a public channel so the company, everybody in the company can ask my bot for feedback and everyone else can see how my bot gives feedback and then can see how I coached the bot and the person on how to think about it differently. And the rate at which context spread has started to happen has gone up like 100x. It is astonishing. It is astonishing. And I was talking to someone yesterday, uh, in Cincinnati who gave me this idea to do the same thing for each function. So, because like there's this, there's this whole thing right now of like, you want to flatten the organization. How do you manage like 80 people, blah, blah, blah. You create robot managers and what those robot managers do is you like create like an account. Let's just take account management, for example, you take an account management channel in that channel, the account managers can say, hey, here's the situation, or like, here's the customer thread. Can you go in and give me a suggestion for how to reply? And because it is programmed to use all of the like best practices, learn what works and what doesn't with the customer and why, and it'll command it directly to memory. And there's no training time other than the live feedback that it's getting, it becomes a highly effective functional coach to every function in your company. There's a guy out there who does AI customer service. It's gonna be Amit, Amit from Richpanel, and he's gonna be like, I have the next thing. I have the next thing. I'm gonna build these AI managers and these AI CEOs, right? Because they have all the infra. Right. They've done this before, dude. Yeah. I mean, yeah, I really recommend Jeremiah. There's like a couple of convos you and I have had about like, cause we talk about like, how do we think about the team and managing the team and bringing the team along? Dude, you should have a bot do the things that you're doing instead. Could the Codex agent could even infer in the channels, like when it should be invoked if you ask it to. I don't want to, uh, say too much because the people that I'm working on this with are like really cool and I don't want to just name drop for, um, for no reason. But, uh, we're working on this thing inspired by what you brought up before, Rashad. Um, this, uh, self-improving feedback loop for a specific purpose for like, um, e-commerce businesses and it learns. Right? It gets better. It doesn't redo tasks. It's, it, I think a lot of it, all of it, all of that component of not just doing the job, but getting better at doing the job over time was inspired by the fact that you said, you mentioned bringing up these feedback loops that self-educate, right? When it does something wrong, it's not just self-correcting for that individual, it's self-correcting for the full function of what that thing does so it can all get better. And I think that there's something in that for me that's really exciting because you're not just the first person to have built something that does something very, very cool, but if you're the first person to do it very, very well, and then it gets adoption and you have 10, 20, 100, 200 users using it and contributing to it, you have something that is just so incredibly robust, um, and always getting better. I, again, I don't know how defensible that is, but it's cool that it can do something cool and get better at it over time. Right? How does the team feel about that? Are they comfortable with the idea of talking to a bot CEO? Dude, it's way better than the real thing, especially for the team. Well, you've solved something interesting as well, is that there's no procrastination. You can't build 'Well, Rishabh didn't respond to me in time,' and make that the reason why you didn't get your job done, right? It's now like, you got your response, go ahead, no blockers, right? And if you can do that in multiple layers, right, then that requirement layer of, 'Oh, well, I need to do this, but I have these questions that maybe should have already been answered before, right? But I have to have them in my bandwidth, in my human context, in order to continue the task and get the job done.' You've solved that, right? Forget actually having like a suitable or better replacement for you to answer these questions. These people are now no longer blocked by needing a conversation with you, right? And that would be my main, the main thing that I would benefit from because man, the last thing I want is a DM with a question. Because, you know, I mean, I answer the question once for one person and another person messages me the same question, I have to answer it again. It's a nightmare. I don't, I don't respond to any DMs anymore. This is just not, it's not scalable unless they're paying me, unless they're paying me, then I have to respond. Yeah. Yash, do you feel like that's, do you feel like, yeah, do you feel like culturally it would be accepted if you said like, talk to the bot, don't talk to me? I think for the size that we are at, because we're like 20 people, I think they expect me to be available. I think if we were like 50 or 60 people, it may be like more acceptable. I think, I mean, like if you take it to the extreme, if you're like 3 people in the company and you're like, hey, talk to the bot, feels a little bit like, why can't you just answer it, right? So I think like we're maybe like a little bit small, but I think like for transact— I see Jack's point. I think for like more transactional questions where it's like, yeah, I've answered this question a bajillion times across like 6 different people already. I think it makes sense if the bot just has access to all all of my chats and can just pull different threads. But I don't know, like, what does that mean in terms of privacy? Like, does it just like learn from everything I say to everyone? And then like, can it just like read feedback and people just ask it for like, hey, what was the feedback that Yosh gave the marketing person? And just like, just reads it out because it has access to the data. I don't know. I don't know how that works. I mean, these things are easy. I think there's a beautiful irony in the name Slack., right? Because if you are responding to Slack messages all day, you're not getting work done. Maybe you're permitting work to get done, but man, the amount of dumb questions you would have to answer all day long if you were just sitting in Slack facilitating DMs, right? Like if you actually replied to everyone who shot you a silly message, I remember Rashad, when your name changed to like, read this document before you send me a message. Like if you facilitated it, if you encouraged it, By like replying to everyone, man, that's, that'd be a full-time job. Just replying to people on Slack, just slacking. I get no work done. Right. Yeah. Yeah. I mean, my current Slack name is Rishabh, please email action items. But I think I, Yosh, I think I would push you a little bit on this. I think, I don't know that people actually want to talk to me or us. Like, I think they want to get good work done. And I think that if that requires talking to someone, in fact, I think I've noticed people are asking more questions with more detail to my bot than they would ever ask me. And I think you should— I can see that. You should, you should check this because what happens is people worry about, oh, it has to be an important enough question for me to ask Rishabh. And there's all of these other shenanigans that show up in people's psychology. So technology reduces the barrier by increasing availability, right? Roughly, very roughly, right? Like you make the thing that was scarce much more available. So if CEO time is scarce, you got to increase that massively. And once you do that, the Jevons paradox thing happens and it just increases the amount that you can. You can be connected to, and so actually people feel more connected to you, not less connected. Yeah, I like that. Yeah, I'm gonna check it out. I guess there's a world, especially in Slack, where your bot can use your handle and use your user, right? Um, so there could be like no separation, but I mean, I guess that's not, it's not super ideal. You just gotta let people know, hey, every now and then it will be me. Yeah, that's what OpenClaw is. I don't, I don't like that. Everybody can have their own opinion. I like the idea that the bot is its own entity. Um, but anyway, that's just my view. All right, Yash, let's talk about the new product. Um, you mentioned that there was like a, like a wider business case for the product, but, uh, dude, this launch video, so good. Have you guys seen it? Yeah. Uh, Jeremiah? No. No? Yash, can you pull it up? Yeah, let's watch it. Did you know 70% of the impact of your influencer affiliates is not showing up on your affiliate dashboard right now? That's because your platform only tracks the last click, which means that one creator who introduced someone to your brand, or the other one who nurtured someone over 5 posts, or that person whose DMs convinced somebody to buy, none of that counts. You're measuring a fraction of your own success. As an influencer affiliate manager, your program is driving impact, 70% of which is not being seen by leadership on dashboards right now. That means, in other words, over 3.5 days of your own work does not show up in dashboards.— that's pretty bad. Affiliate marketing is riddled with these issues nobody is talking about. Your links and codes are on Reddit as we speak. Every time a customer Googles a code at checkout, you end up paying commission on a sale that was anyway gonna happen. Your best creator has posted about you 5 times last month, but a last-click affiliate who drops the code at the right time gets all the credit for the sale. Every quarter, your CFO asks, "Is this spend incremental? Are these sales even real?" And I know that your gut says yes, but your dashboard says no. You're constantly following up with influencers, chasing for W-9s, billing invoices, following up for content, and doing all of this boring work that is not marketing. This is not why you became a marketer. Meanwhile, your affiliate platform, while they should have solved this years ago, they're all stuck on last-click attribution and dashboards that look like they were built in 2015. You deserve much better, and that's what I bring to you today. Over the last 3 years at Xerall, we've helped e-commerce brands build influencer and affiliate programs, driving over $59 million in revenue. And we kept seeing the same problems. And that is why we built Saathi, the modern e-com affiliate platform for the modern world of influence and agentech commerce. It transforms affiliate marketing in 3 magical ways. We have cookieless multi-touch tracking, which is a breakthrough because it works even when people have cookies disabled or ad blockers enabled. And it's multi-touch, meaning the creator who drove awareness on TikTok still gets credit for the sale when it happens weeks later. Next, there is powerful AI fraud handling, which keeps your program safe and honest. Saathi flags the basics like self-referrals or leaked codes, but it goes beyond and tracks if someone's hijacking paid traffic and spots fishy order patterns. And finally, we've created a beautiful affiliate experience that's gamified and makes affiliates wanting to post more and make sales for you without you following up with them. All of that, plus there are AI automations that take care of all the boring taxes, payments, follow-ups, reporting for you so you can focus on being a marketer. Saathi is your companion that takes care of everything else so you can be a marketer. Feedback from our early users has been unreal. Me and the team can't wait to show you the hidden potential in your affiliate program with Saathi. Welcome to the new world of affiliate marketing. I will see you on the other side. Wait a second. Wait, wait, wait, wait, wait. Did you put a copyrighted song right at the end of your launch video? I did. Yeah. LinkedIn has very loose copyright laws I found. So, you know, I figured might as well. Very cool song. Yeah, it sounded really cool. Yeah, that's awesome. Yeah. Very cool. And how's it been since the launch? Been crazy. Like, yeah, not 20 hours, but definitely like 15-16 hour days on this side, lots of demos. And we've done an interesting like waitlist strategy. So we almost like, if you guys remember the Superhuman launch where you had to like get on the waitlist to get in and then you could be like bumped up if somebody used your referral code to like also sign up and things like that. So we're executing that. So we're like curating how we're onboarding brands and that's actually going well for us. So like, yeah, lots of demos. And like I'm obviously like personally involved in everything because it's like brand new and we're like figuring everything out. So just in founder mode, but yeah, gone really well. I think it has— this video on LinkedIn has over 12,000 views or something like that, just on LinkedIn. So I think everyone in this space pretty much, like brands, have seen it and commented on it and so on. So it did like— we put in a lot of work, but I think like it just performed really, really well for us. So yeah. Did you hire somebody to help with that or did your team do all of it yourselves? The, no, for the, for the shoot we did, yeah, we hired some, like some friends of mine who went to film school, but the, the direction, the, like the script and everything was, was all internal. Nice. Yeah, it's really good. Yeah. I wanted to ask you about this. Uh, it could be a good place to wrap up on. Um, you mentioned founder mode and actually being in the sales calls. Rashad just mentioned jumping off to go to a sales call. And we were just talking to Jacob Postle, who's just launched HQ, and this is his calendar. It's like back-to-back calls. I think he posted like a post onto his page, like, how do people raise and deal with a calendar full of sales calls? And I was like, yeah, I read it and I thought, sales reps, right? You hire sales reps and then you go do the thing that only you can do. Oh yeah, this is it. The paradox of fundraising. Founders with bad businesses run a full process pitching a gazillion VCs. Founders with good businesses busy running a company. My calendar is 10 hours a day of demos. The rest I'm talking to customers. I thought this was what sales reps were for, but like, obviously there's this founder's mode and like, you know, I'm in, I do all of our sales calls because we're tiny. But yeah, I'd love to hear from you guys. You know, Jeremiah, are you still in sales calls and Yash, what's your, what's your thought process to this? I mean, I can see this paradox, but I can also see like, you know, I've, a few of my friends have software businesses and they have sales reps. They're not in any of the sales calls. Right. Yeah. I mean, most of our customers, we have a couple of different types of, um, revenue streams, but majority of it is, uh, you know, software subscription, somebody signing up to use the core piece of our, our platform. And, uh, I think our largest customer is like 0.6% of our revenue in that context. And so I just don't do those calls. It's just like, there's no, there's no customer I'm going to talk to that's going to move the needle in a way that's worth me getting on a call and spending a bunch of time on it. And in that context, now I'll, I'll communicate with people and, you know, send people tips and tricks and all that kind of stuff, but I'm not going to do a sales call. And then we have a couple of other things like, you know, we have some relationships where it's like bigger bulk pricing type things, where it's like a relationship with an agency or relationship with an ad platform or whatever those things are. And that's a different story. But you know, when you're, when most, there's only a handful of those and there's like, you know, couple opportunities a quarter like that, that I engage with. And otherwise I just leave it to the sales reps. It's just not. We're at a scale now where that's, it's not worth my time to, to be on that. Uh, I feel, I hate saying that because it's not that I don't care about the customers, it's just like I literally don't have time to spend with thousands of customers. It's just not an option. Um, for me to give that to everybody, but the first like 3 years I was basically doing all the sales calls, 2, 2 to 3 years. Um, and then ultimately just kind of handed that off over time. Yeah, I like that. I think we're— I'm kind of seeing both play out at the same time because we have Serral, which is kind of like our older legacy company, which is still running, still active, still onboarding customers. And I'm not like involved, like we're now at a stage, like, I mean, there used to be a time where every single customer we onboard, I knew them by name, I knew the city they lived in, like I knew what their hobbies were, like everything. Now we're at a point where we like our head of sales, Max, like he just makes sales and we just know like, and that just works, right? So like I'm not involved in that whatsoever. Even on the customer success side, there's Chris on the team, like he takes care of it. Like I'm not super involved in that. I'm looking at metrics and like sometimes like I get looped into a sales call, but maybe it takes maybe twice a month, right? Which is like maybe an hour of sales calls on the Saral side. On the Saathi side, because it's like, like literally today is like one month as of the recording. It's like brand spanking new. Everything, I'm involved in everything, like talking to customers on the product side, sales calls, onboarding calls, setting them up, migrating their affiliates, like I'm in everything. So I think like I get to see like both worlds where Searle now is almost 3 years old. So that's like set up and running. But Saathi, because it's new, I get to do it almost all over again in the right way. So almost like I'm like a second time founder in like the same company in some sense, you know, so yeah. Awesome. And how fast are you trying to get out of sales calls? And I guess the takeaway for somebody who is in sales calls every day, you know, yeah, when should you be getting out of sales calls? Obviously at your bandwidth. But like Jacob, to some extent, I imagine people want to talk to Jacob and that's why he's doing 10 hours of sales calls a day, but that can't scale, right? Like if you, if I imagine like even if it's a week out, there's opportunity cost there. Or like if people can't get on your calendar on the same week, then are you losing customers just because you're not hiring a sales rep? Yeah, I think the way I would think about it for us, because we're new, is like rate of learning. Right now on every single sales call, I'm like learning something new. I'm taking notes, I'm passing it on to product versus like on Searle, we're not necessarily like learning those many things on sales calls. We learn things when we like talk to customers who are deep and they're like so at the edge of their program that they want this one thing that the platform doesn't do and we need to build that. That's where we learn. But I think like it's like how top of funnel is the learning, right? I think like for Saati, it's like at the sales call level. For Searle, it's like customer success once they're activated, what do they need? It's like, okay, the customer is like 6 months old is when I talk to them. I think I look at it as I think Jeremiah is nodding, so you probably agree. I look at it from a rate of learning perspective. Like once I, once like the rate of learning, however you want to measure it, like insights per call or whatever, like once I feel it drop off, I think it's, I feel more comfortable even giving it or delegating it to a sales rep. Saying, okay, like, yeah, you just do it. And then I, I go like lower and lower in the funnel because then like there's more learnings to be had. But like, I think of myself as like context learning machine because I want to be like an expert in how the customer thinks and what they want. Yeah, I think that's really interesting. The, um, and I would agree with that. I like the rate of learning framing of it. I think of it kind of in terms of like product market fit, which everybody has a different way of defining. But it's similar concept of like, do you, once you know that the product meets the customer's need at whatever price point, then at that point, that's really, that's the first point that you can really hand it off to somebody who isn't a co-founder level person. Maybe that's a little unfair, but anyway, you need the authority to make decisions on the fly and to be able to say, okay, this doesn't work for a person. So let's try something else. And you're basically experimenting, experimenting, experimenting. So once you kind of start to get out of that experimentation phase and you have something works, that's when I would hand it off. Um, and that could be a month. It's probably like a year or two. Yeah. You know what's really interesting about both those answers though? Neither of you spoke about selling. You both spoke about making sure that the product is right for the customer, right? And obviously the salesperson comes in with that as the main objective, right? Like sell. But like the, this is a different approach. It's, I want to make sure the product is right for the customer who's buying. And get the sale that way, right? Yeah. Interesting. Yeah, because I mean, at that point, ideally you don't have to solve that problem for a sales rep. It's just like a matter of them identifying good fit customers and getting them. Yeah, because like our GTM is so new, I get to like learn what's working. Hey, did you actually see the launch video? Did you see like a customer post about it? Or like, how did you come? So like, I'm learning that. If I want to make like pricing changes, I can just do that. Like, hey, okay, there's like this component here. Okay, you get this feature here. Like, I'm learning all of that versus if I give it to a sales rep, then that like, so they have to like, okay, they gotta like talk to me about it or like, you know, somebody about it and then they change the pricing and then it's like, it just slows the whole process down versus when I'm in it, then I can just take these calls and then, so again, this is kind of like pricing market fit or whatever you want to call it, right? So yeah. Cool. I think that might be a good place to finish it off. It was fun, guys. I learned something very good at the end there. I mean, I always learn so much in these calls, so yeah, I appreciate you coming back on, Yash. Yeah, thank you. Thank you for having me again. Yep.