Episode 34
A conversation with Shivani Poddar
Published · 58 min
Shivani Poddar led engineering on Google DeepMind's autonomous coding agent. Then she went solo to build Fo — the personal agent whose viral launch admitted what no AI lab will: AI often can't finish the job alone, so her agent hires vetted humans and picks up the phone itself. A decade of building assistants at Alexa, Meta, and DeepMind, distilled into the sharpest product thesis of the agent era — plus her playbook for building a founding team without a co-founder.
What Shivani talks about
Pick a question to ask it of this episode. Answers come from the transcript.
Transcript
Transcribed automatically, so names and words may be wrong. Each time opens the video at that moment.
0:00Shivani Poddar: I've been building personal agents before. They were called personal agents for quite some time. I built the first Facebook assistant, which was back in 2018. The reason I left to start the company was because I wanted to take a few risky bets. I wanted to give the agent a phone. When we started our phone call, hang up rate was like 80%. Now the hang up rate is between 20 to 30%.
0:21Julian Weisser: What's the bear case for solo founding?
0:24Shivani Poddar: The founding journey never really gets easier. You definitely, as a solo founder, have unique stresses that you have to carry alone.
0:31Julian Weisser: What would the case for being solo be?
0:34Shivani Poddar: What the AI worlds of the last few months even have afforded us is genuinely an opportunity that has not existed. I'd never be against having a co-founder, but I'd never rush into it just because I wouldn't want to do it solo.
0:49Julian Weisser: What do you think it was about the announcement video that caused people to, you know, really light up? Because it could have just been something where everybody said, oh, it's just another one. Yeah. You know, but but that didn't happen. And what do you think it was about what you did that caused people to actually have that reaction?
1:06Shivani Poddar: Yeah. First off, I'm so excited to get to know you in the show and solo founders being a solo founder. Like, I hadn't discovered this community soon enough. Um, I think, uh, when I think about what sort of clicked for us is there is almost this, uh, contrarian view, which is funny now emerging where on the one hand we say AI is everything. AI will take our jobs. And, you know, autonomous AI is here to rule. And on the other hand, at Fargo and in building four, we've taken this approach of, um, yeah, I may not be just I may not be the only thing. I might also need help and collaboration to truly bring magic to people. And I think that's really what I like to people, because all of us have had this experience where we're using our favorite personal agent, you have whatever, and then suddenly the agent gets stuck in the work falls on you, whether it's, you know, solving a Captcha. And I joke, I'm like, if I want it to be a captcha monkey, would I get an agent?
2:16Shivani Poddar: Or it's like, um, you know, pick up a phone or negotiate with somebody or actually be, you know, have some, just have really good taste in something that I don't have good taste. The the possibilities are endless. And I could almost sense that people had that fatigue of going to these agents with this dream of just giving it things to do and not finding that it goes all the way. So I think that's probably my hypothesis on what clicked that. And then I'll maybe say something else, which is, uh, people saw themselves in just me and my team. Um, I think it's funny because do you remember this tweet was trending? Somebody was like, oh, people keep billing me the Silicon Valley husband where I want the Silicon Valley wife.
3:03Julian Weisser: I think there was Sonia who posted that.
3:05Shivani Poddar: Yeah, that was like such a good representation because like, when I'm building these things, I'm just coming at it from such a different point of view. Because just to your point, like just the life experiences have been so different.
3:18Julian Weisser: I think that it's very interesting when oftentimes in tech we have this, uh, this dream of what the future could look like. You know, in terms of what technology could enable. And then oftentimes sort of the V1 kind of falls short of it. And I think that what you're describing is sort of this disenchantment with, you know, the allure of the idea, getting nerd sniped by it. I think that this is often the case with things like OpenClaw, for instance, uh, where people really love the idea. Um, and then some people, some subset of those people really enjoyed going in and sort of trying to tinker with it, and I think that it was. It's amazing for tinkerers, but I think that a lot of people don't actually want to be tinkerers. They want to use tools that that are reliable and that work well. And I think that in a lot of ways, a lot of the current tools fall short. Also, what you described is this idea of, I'd love you to expand that a little bit, this idea of kind of you hand some task off to something and then you expect the next thing to be being done, but instead the next thing is that it didn't actually do what you all did or or it fell short. And now you have to kind of remediate. In a lot of ways, doing that remediation work is just as tiring, or even maybe more so than just having done it yourself in the first place. Can you share a little bit more about kind of, um, kind of how how you thought about this as you were sort of scoping out maybe building building a company or was that was that was that sort of what happened as you were actually going out to try and build a company? Or did you kind of start from the premise of, I want to try and work on a company right now. And then ultimately this is kind of where you ended up honing in.
4:56Shivani Poddar: You know, this is been my I think this has been a product in the making for the last ten years. I almost want to say, um, my so I actually wouldn't have started a company if this wasn't the conviction that I had. So back in and, and the conviction actually started bubbling in even before OpenClaw came out. OpenClaw was actually such a good wave to know that, you know, people want this, and yet it's going to be so hard for anybody outside of, you know, our Silicon Valley community to have this. So is a very good, um, validation mechanism. But I've been building personal agents before. They were called personal agents. So quite some time, um, I built the first ever social chit chat bot that was deployed in Alexa back in 2016. And then I built the first Facebook assistant, uh, which was back in 2018, which was a device called Facebook Portal. And I remember actually handwriting the smart calling logic, which let the assistant call your loved ones from the device, which is speech to text. Um, and so it's been something I've been added for a while. And, I mean, the reason I left to start the company was because I wanted to take a few risky bets. I wanted to give the agent a phone. You know, if you're feeling, just call somebody. About 40% of the world is not digitized yet. Believe it or not, it does take us to pick up a phone to make that thing that. And then, uh, it's funny because, like, when the agent does pick up the phone, there might be a human on the other side who does not want to talk to an agent. And so I, I just I just had this dream of, um, if I, if I thought about the most competent person in my life and captured them into technology, like, what would that be? Just the trust that not only will they share my private information without asking me, but trust that if I get give them something to do, they'll just get it done. Um, and so that's sort of the thesis that I've had from the get go. And so we built the technology. You have to. Otherwise, it's really hard to retroactively adapt these things in a way where they can collaborate. We've had to teach models to collaborate, and they're not trained to collaborate because they're trained on data, which is like copilot. Um, and so we've had to teach the models that they're failing because when you know that you're failing introspection, then you know that you need help. And then can you understand where you get this help from? And over time, you know, build this almost like a team like dynamic with your AI human team. So, yeah, it was very much from the get go.
7:37Julian Weisser: Hey there. We hope you're enjoying this episode of the Solo Founders Podcast. If you haven't already subscribed, you should make sure to do so so that you can get every episode as we release them each week. And of course, we always appreciate it. If you share the solo founder's podcast with friends, it's what causes this show to grow and help more people. If you are already building a startup or considering building a startup solo, you should apply for the Solo Founders Program. We bring solo founders to San Francisco to build solo together for three months. We work closely with each company and we invest $100,000 as well. You can learn more at Solo Founders.
8:16Julian Weisser: And without further ado, let's get back into the Solo Founders podcast. I'm really curious to hear, as you think about sort of the evolution, because I feel like there are very few people who've gone on this path, uh, starting, I think you said almost ten years ago you were working, uh, with, you know, with agents, sort of, or at least like the precursor to agent tech software. Um, you've covered the bases with Alexa, with Facebook now. Meta. I'm curious. I'm curious kind of what you've seen in terms of the capabilities and kind of what caused you to realize we were kind of at this tipping point. I know you mentioned OpenClaw, but what were some other things that kind of made you feel like now was the right time to be starting a company beyond OpenClaw?
8:59Shivani Poddar: Yeah. Well, I was actually at Google before I left. I was at DeepMind first, and then I was at Google Labs. And actually what we were building at Google Labs was this, um, which was actually, at the time, maybe the first one. Juelz, the autonomous code agent, which was living in a cloud, came with its own computer. And I kept thinking about the fact that this agent could dream. This agent could be learning about your code base while you're asleep. And when you wake up and this now feels like, oh yeah, duh. Yeah. But at the time, the industry was at Ides, we were at copilot. And so I think that was the tipping point. When we were building jewels, I realized that, um, the the journey between autonomy, proactivity and deep context is the ability to design cross organizationally. Cross. Um,
9:53Shivani Poddar: cross channel almost to make this possible. Like it was no longer, um, it was no longer only a model capability issue that needed to be solved.
10:04Maybe that. Do you think.
10:05Julian Weisser: About. I'm curious what your philosophy around the phone, because I think that you're right that in a lot of ways the default is the agent says, would you like me to email? You know, would you like me to try and find an email account? Yeah. And oftentimes there might be an email, but that's usually the worst way to get an answer or anything. You should just call. Yeah. And in fact, I think this is probably true in a lot of cases, even when you're just trying to talk to people who aren't small businesses. If you can get somebody on the phone, you can usually get there much faster. Um, any any interesting learnings about about phone calls and things like that and that dynamic?
10:38Shivani Poddar: Oh my God. Um, I like initially when we started our phone call hang up rate where people would listen to us and like, hang up was like 80% super high. Yeah, it was super high. Um, now we're we're like at the hang up rate is between 20 to 30%, which is insane, because if you think about phone calls, it's very interesting. It took me a second to realize when you're calling somebody and there's an AI on the other side, you have an incentive to stay on the line because you have a mission to accomplish. Now what? When you're getting a call and you have no idea why you would be on a call with this thing, you have no incentive to stay on the line. And so for us, we had to almost design. Well that's one. And then the second thing is you do have regulatory postures. You have to mint. You have to disclose. It's an AI, you have to disclose. It's a recorded line, which is scary for the person on the other side to listen to. And so we had to like think about, um, three things really well. One is just designing this experience in a way that the agent doesn't get hung up on. Two is, um, the stake for our agents to sound natural, to have tone taking, which just feels flowing, um, to use the right kind of data. Not like giveaway data is insanely high, much higher than a customer support agent. So we had to think about the technology much more deeply than just plugging in the next voice model that we got. It took us a very long time to build that. And third, actually, um, you'd be surprised when an AI calls somebody who can be, um, on an edge of a bad actor. They can try to trick the AI in giving up information about the user. And so we had to think very deeply about agent disclosure, about, um, personal data sharing and a bunch of these considerations which we now see. Um, can be can be very offputting for a lot of people. So yeah, those were the three things we really had to think about when doing the voice agent.
12:49Julian Weisser: At what point? I mean, it seems like it's already happening. Um, a, a large amount of this seems like, you know, putting aside the bad actors, of course, there are bad actors everywhere you go, but, um, but they are the minority. Um, you know, at some point it feels like there's some aspect of, uh, of sort of it's, it's foreign. It's, it's it's an un sort of unfamiliar thing to get a phone call and say, hey, it's an AI and it's recording and all that, that puts, um, some pressure there. Um, it feels like a lot of it's just like that weirdness. So once where this goes away, it becomes natural. And then also the, the belief from the recipient of the phone call that this is actually legitimate and that they aren't going to have to deal with the really annoying thing that you're used to, which isn't AI, but it's that sort of you call a customer support line and you say something and it just sort of loops back and it's just terrible experience. So I wonder how much of that is exposure, uh, familiarity and realizing that this isn't sort of this the old way of doing things that was painful and nobody likes, uh, maybe that maybe there's just some threshold to get through it. It sounds like that's already. We're already starting to see that happen a lot.
14:00Shivani Poddar: In fact, it's funny because once a business has had business with RTI calling them, um, they will always pick up and want to engage. It's funny. Um, an insurance company from Australia found me on LinkedIn and wrote to me. And this insurance company is, you know, 5 billion are big, so not small by any means. And he was like, the reason I found you is because your AI just dropped insurance on my business. And I realized that your eye is way better than anything we had. And so I had to play with the product. And, um, I'd love for the like, you guys to make sure that your AI keeps finding us because this was the most seamless, you know, interaction on both sides that we've ever had. And so that's the power in that, uh, companies that are leaning in, businesses that are leaning in to this new way of commerce and transactions are actually finding that they're getting discovered way more. Um, and so this is like a big business. I also had a makeup artist who found us and reached out to us, and she told us that she used to be she was like, where is this for listing how? Like so in her mind, she was like, oh, there's probably a new Google somewhere. And so she is like, I like she had listed her business on Google. And so she was like, this poor thing keeps finding me and it gets me the best clients. Like she was behind the scenes on a big AI summit. Uh, just because of this. Then she actually did make up for, um, some really, you know, well-to-do person in the Valley for their wedding or something like that. And so she was like, how do I, like, go and add more details to my faux profile. And so her thing was that she, um, actually any time forward, email her or call her, she would be on top of it. She would make sure she gives all the information. And so actually, she started beating a bunch of her contemporaries who only did SEO or whatever on Google. And so I think there's like a wave emerging where people who are leaning in actually have, for the first time, maybe in many years, to just be more relevant than all the incumbents in their respective industry.
16:22Julian Weisser: At what point, uh, you know, maybe, maybe this is a foreshadowing. At what point will people's Fo be able to like answer the call for them and your businesses. Answer the call for them. You know.
16:32Shivani Poddar: Some people are doing this already. Um, small number, but um, a lot of times people, uh, that's the beauty I think you you mentioned when I walked in, you were like, hey, I see this for personal agents, but, you know, certainly for solopreneurs or small medium businesses that don't have as much means to be on the phone the whole time or a back office. Um, this is useful. And I think to some extent, like there's many companies that are out there doing it, so some people will use those. A lot of people just use for, um, to deploy, you know, a back office, um, small medium businesses. Again, this is something we've not had, um, enough bandwidth to even focus on, but there's so much appetite.
17:14So let's take.
17:15Julian Weisser: Us back to sort of the very beginning of you starting this company. Um, you know, I think that, uh, everybody has different motivations for starting a company, different sort of timing scenarios. Uh, some people are really just keen to start a great business. Some people are keen to start a specific business at a specific time. Uh, you know, surprisingly, I kind of would have assumed that people who were starting a specific business at a specific time would be more likely to be successful. But I've seen enough cases of people who just said, I want to go and start a great business, and I'm not going to stop until I succeed. Do well with that. It sounds like much more, uh, on your side that you had a very particular view. You felt the timing was right. You saw this moment sort of emerge where people liked the promise, but they were let down by the execution of the other alternatives out there. Can you share a little bit more about kind of what what that sort of moment looked like where you said, okay, it's time I'm, you know, I'm leaving. I'm starting a company. What did you do next?
18:09Shivani Poddar: Yeah. This is such a good question because it's taking me back. Um, I, I mean, I think like, um, for me, it's less about, like, building the specific product. It was more about, uh, a mission I am obsessed with, uh, making sure I do more with my time. Um, and so it's funny. Um, this is recording, but I will never share it anywhere else. I have, like, um, an actual hardcover journal, which is like a time productivity. How do you manage your time? Well? How do you optimize the
18:48Shivani Poddar: parts of your life in a interesting way, which is like on Amazon somewhere from years ago. Um, and so I've always had this mission of, um, having more time on your own terms gives you more access to opportunity. And by more time, I actually don't mean, um, those simple productivity scenarios. I actually think that that that sort of like, is already things that, you know, we were able to do with a lot of these tools, but actually mean, like sitting on customer care calls to, you know, get that airline thing fixed for two hours or, um. I used to see my dad try to get my mom's health insurance sorted for her medication, which would take him, like hours and time away that he could be spending with her. So I really mean, like, the nerdy things outside of, like, the usual that it's great to have, like a daily summary of like what your day looks like and whatnot. And so the mission always was that how do you make life more meaningful for anybody out there in the world, not just for the billionaires of the world? And so I think for me, it was the I saw a few different ways in which this mission can be realized with the technology that we have, which sort of led me to go out in the world and do it. Um, and then I think, um, I had to just have conviction, um, that I can actually execute as a solo founder. Um, but also, I think, like it took me years of working at all the big companies to realize that I actually have built the muscle that it takes to go execute. Um, so yeah, that's that's sort of the thing that, uh,
20:29Shivani Poddar: brought me to do it.
20:31Julian Weisser: We'll talk more about this although founder thing in a moment here, but I'm really curious to hear your perspective on, you know, what do you what do you look at? People talk about examples of technology, especially with agents or personal AI. Um, it's often this sort of
20:48Julian Weisser: way that I don't think the general public actually would use AI. Um, and the way you described it there, I think is, uh, is is much, is much closer to what I imagined people would actually want, which is, hey, do you really want to wait on you? Like, it's actually cruel that we make, uh, we make people wait for two hours on a phone call on hold. Or especially when it comes to health care and health insurance and all of that. Like, I actually think it's it's it's a cruelty, um, that, that, you know, business is, um, you know, put on on folks. Um, and, you know, I think that that's that's the sort of thing that's like being an advocate for, you know, for the average person to be able to actually do this and that, that's even more valuable for people whose time is valuable from a monetary perspective if they're currently working or they run a business, but it's just as valuable for somebody who's retired, you know, to somebody who's retired, really want to spend their retirement, you know, in on hold with, you know, insurance companies. It's crazy. Right. So I'm curious why why do you think it is that so many other folks, when they talk about uses of AI, they always seem to be something that's like booking a restaurant or something versus something that's so valuable. Uh, around like the the doldrums and sort of the pain, uh, associated with, like, life.
22:08Shivani Poddar: Yeah. I think like, there's, there's a couple of ways to think about this. And I'm curious your take because I'm sure you've played around with these agents as well as much as I have. For me, I think the reason. So I, I wanted to build a company because I had this thesis that this can help the, the average Joe out there. And so actually what I did before we actually did our public launch, um, instead of giving early access to just people in the Valley because, you know, we all will use as many tools you.
22:40Give us at all.
22:41Shivani Poddar: Yeah. Where, where we have no loyalties to any tool. Um, I actually decided that I was going to find a way to see these in the suburbs or in contractors in Ohio or, um, somebody in India, and just see how do they actually use it and figure out then where are the gaps of what they want to do and what they're actually able to do with that? And I think that was a very big eye opening experience, because none of these people want to book a restaurant with it. A lot of the people who are even transacting or trying to plan a holiday or have kids and are trying to like, do a lot of wrangling. Restaurant is like the simplest thing that they might even do with, like, you know, your existing tools. And so, um, it I realized that while it's the easiest way to explain this technology because it's single hop, it's simple to say, you can book a restaurant. Everybody understands it. It's much harder to explain the tougher use cases where coordination needs to be had, where it's multi-day, where it's multi-party coordination. And so I realized that one, the normal people actually don't use it to book restaurants. I mean, even if they do, that's not the one that they're coming back. That's not the use case they come back for or that's not the use case. Why they feel excited to be using just this product. Um, and, uh, at the same time. Sometimes some of these use cases are the ways in which people understand the technology the easiest. Maybe.
24:19Julian Weisser: You know, it's funny because I think that the other the other thing is like the the amount of benefit you can give someone to not have them sit on a two hour hold is so much higher than the amount of time to phone. If you actually are calling a restaurant to make a reservation. The amount of time that actually takes a person to do is relatively low. Um, so having an agent do it for you is is nice, but you're also probably having to communicate to the agent what you want. Um, and obviously the agent might be able to infer, but at the same time, the amount of, of sort of work that you could have an agent do for you talking to your bank or talking to, you know, you know, doctor's offices or something like that feels feels like it's much better. I'm curious, what are some things that you've seen that have surprised you about? You know, what what folks are are using Fo for now.
25:08Shivani Poddar: I've been surprised, honestly, by everything I see. And so I'm gonna try to pick out some of these scenarios. But even for me as a user, um, what I use it for would just was just the iceberg of what people have started putting this on. And so I think, um, some of the big ones, uh, people love having pho be the social agent in their group chats. Um, and I think that there's a reason for why they do that is because, um, there's almost a sense of safety that it's not gonna, you know, come, uh, share any of my secrets from my personal chat in this chat. And then, um, these are typically the younger audiences they will send for to apply for jobs for them. And job websites are very interesting because applying for job could be over LinkedIn or over email or over websites. You don't have to call for jobs, but, you know, outside of that. And so we see a lot of these scenarios where it's relatively hard to put together an interesting job application. So sometimes when our humans in the loop come in, they actually teach the agent how to put a better profile together for people. So that's been one. Um, healthcare has been a big one. Um, where where, by the way, already pursuing SOC to, um, our stack is compliant, our stack is HIPAA compliant. But I do want to get those certifications because I think that healthcare is game changing for people, whether it's calling ten doctors offices to schedule an emergency colonoscopy, which somebody has done, or a man in Florida actually use to actually fill out his Medicaid application, which typically takes weeks, and a social worker. And in this case, you know, we actually took the strain off of the social system. I just for having done it. Um, and then I think it's also some of the fun ones, I think because, uh, follows. Been built by somebody who is a woman who's a founder. I think I've like really thought about things like what was great at shopping in that I if I send a photo of your outfit to any other agent in the market, guess who's going to nail it?
27:22Hmm.
27:24Shivani Poddar: Including finding discounts.
27:26Okay.
27:27Shivani Poddar: And so there's there's obviously commerce use cases, which people use us a lot for. Um, and I do think that another interesting thing that I've realized is not just commerce here, but also international commerce is something people really trust us for. Uh, could be because we own our own payment rails. So we're not restricted to just us only shopping, but people like to shop internationally with us a lot.
27:54Julian Weisser: So I'd love to just talk about kind of where you were and sort of the beginning and understand a little bit more about kind of what this V1 actually looked like, because ultimately, um, before you actually launched publicly, there was a lot that you could do. This was one of the most sophisticated product launches. The reason that people were so over the moon about it was because of just how complex the product was in terms of not not in terms of the complexity of using it, but the complexity of what it was capable of. Um, you know, it's we're in a different era where it's hard to imagine, um, you know, in a different era that you could have built something in a year or under a year and get to the point that you have with Fo. But what is the very first version look like that you shared with somebody who wasn't, you know, a team mate or, you know, a very close friend?
28:42Shivani Poddar: That is such a good question. Um, I mean, first off, I think, uh, even as a solo founder, one thing that I think is incredibly important outside of having the best AI tools is having a very strong team. And so the first thing that I knew I needed to do was actually twofold. One obviously started prototyping what this product was going to look like, and it started with a very interesting mindset. And I was telling you about this before, but the idea was that this is going to be the product that helps me build my company. As a solo founder, I'm almost like this prosumer who can use it from both sides. And if the product fails me, then the company doesn't have a right to be successful for this product to be used by anybody else. And um, as a solo founder, my biggest use case at the time was scheduling. And so the V0 absolute first thing that I built, and it was very interesting because I when I remember that stack, it was not built for scale at all. It was built, uh, it was built on very interesting first principles, which we still have, which we still hold, which was it was going to be remote. So it wasn't going to take my local data and open it up to the world. So not the open floor architecture. Um, and the second thing was, it was gonna be empathetic. It was gonna really understand what the user needed and thought it was going to be reliable. So for even things like scheduling, it was going to be reliable. So that was the V0. I built that and I actually used that to do all of my recruiting, all of my investor meetings, just to make sure that if this fails and it it had a bunch of hiccups and I learned those and I knew that those mistakes weren't fatal because, you know, we kept going. Um, it was if it failed, it wasn't going to make it. And so that was the first thesis is taking something which was very repeatable, very high value, but also very high risk, because you put your product almost out there and using that to make sure that you build a very, very high functioning team, I think was my first way of being the solo founder in this journey.
31:04Julian Weisser: That's really cool. And maybe you could share a little bit more about how you thought about early team because, you know, solo founders, um, you know, they, they essentially because they don't have co-founders, they usually hire earlier. Yeah. And they also think about founding team differently. Yeah. So I'm curious, what is your philosophy been about, you know, building out this incredible founding team and what, what how did you think about the first hires? Like, what were the first hires that you had join your team?
31:30Yeah.
31:31Shivani Poddar: Um, I mean, first off, even at company fundamentals level, um, I made sure I have a really big employee equity pool, because if the company goes up, my first team should feel like we're all going to the moon together. That's been a core principle. And so even for other solo founders, as they set up their team, as they set up just the structure of their company, that was incredibly Important. Um, and so that was like the first sort of foundation. And then the second thing I wanted to do is I did not want to cold hire. I only wanted to hire the best people and the people I either knew directly off or I knew off of. And so, um, the first few people I hired, um, amazing, like, incredibly instrumental to our success. Um, I was actually a jewels, and I hired a few people from the startup, which was Aqua hired at the time. And I knew that those people were really smart, and I knew that they had other people from that company that were really smart, that may not love working in a big company, but would love working in a startup. So when I left, I was like, okay, I know this general set of people who like really, really smart and can do really good work. And I've worked on autonomous code agents. And so I went out and I found all the insane people that I wanted to hire at Google, but I couldn't because they didn't want to work for a big company and I was like, guys, we're building a rocket ship. And it's funny because if you ever see my messages to people that I sent to hire them, um, one, it's all handwritten. I never have for send the first message for takes on afterwards, but it's legit. It's three words. Want to join a rocket ship with a rocket emoji? That's it.
33:19Are you good.
33:19Julian Weisser: At for right that. If that's.
33:21All.
33:21Shivani Poddar: It's just, I don't know, it's.
33:23Maybe just, uh.
33:24Shivani Poddar: I don't know, maybe it's just, uh, you know, it's just a touch thing, like, if I've made that effort, like, I really care about you coming on board.
33:33Julian Weisser: Yeah, yeah, I think that something that would be very interesting. A lot of folks who listen to this podcast, they work at Big Tech and they're considering, you know, leaving, um, and starting a company. And I think that, you know, lots of people can leave these bigger organizations and start amazing companies. But I think that one thing that they're always curious about, and I think we've heard a lot of perspective so far. I'd love to hear yours is how to think about when you're having conversations with people.
34:00Yeah.
34:01Julian Weisser: To understand whether or not they will make the jump from, you know, being an incredible contributor at a larger company to being sort of an early, a successful early teammate at a startup. I think there's a there's a question about that when it comes to founders who are coming from those places. But I think it also applies to early team. And how do you how do you think about that? What do you what kind of questions do you ask people to understand if this is sort of the right, the right sort of stage and the right type of company for them?
34:30Yeah.
34:30Shivani Poddar: This is such a good question. And honestly, because there's incredible talent at all of these places, which anyone would be lucky to have. Um, at the same time, to your point, like a lot of the talent might actually be the worst person that a solo founder could hire. Um, I think, um, for me personally, when I started my company, I
34:54Shivani Poddar: had almost internalized that maybe the first 2 or 3 people should be people who've actually been founding engineers before, just so they know how. And the journey is insane. Yeah. Um, it's almost like, don't do it unless you're ready for insanity. And so the first few people were a rocket ship, like, who has done founding engineering before and, like, has been really good at it. Um, and right after that, when I started hiring from
35:25Shivani Poddar: all of the become, we have incredible people who've come from Gemini CLI and a bunch of these other places. It was about who is a tinkerer who has either been attracted to 0 to 1 products in big Tech sometimes and frustrated by it, but like keep kept going at it. Um, or who just has side projects that they can't stop doing because they just love ideas, curiosity and building. Um, building, you know, almost tangibly building. And so I think those were my singular two most important things to think about, and probably also things that now, when I think back about my journey, were true for me, I was always in the 0 to 1 products, I always had new ideas, and I was always building something on the side, even if I couldn't do it as my day job.
36:18Julian Weisser: That's really that's really interesting. The I'd be really curious to hear we talk we touched on a little bit earlier, but this, this perspective of human in the loop. Um, having having humans like do calls and and do work with the AI. Um, obviously this is more cost intensive, um, than AI, but at the same time, this is the thing that actually causes AI to not fall back and sort of not live up to its promise, or not even AI to to a service. Because if you think of what what pho is, it's sure there's like a large AI component. But what it really is at the end of the day is the service, the service. And and it's it's, you know, maybe it's 80%, 90%, 95% AI. But the but the thing that gets it over the last mile on maybe some of the hardest tasks or tasks where maybe the people who are getting phone calls from AI don't want to talk to AI. The thing that gets at that last mile, at least right now, is, is humans. And that also makes the AI better over time. Um, so I'm curious sort of
37:19Julian Weisser: when did you when did you sort of come to that conclusion? How do you actually do that? Do you consider yourself to be somewhat of a market, like a managed marketplace? How does how does that all work?
37:28Shivani Poddar: This this took a lot of thought and a lot of effort because we were we wanted to make two things right as we were designing the system. One, we didn't want it to feel like a throw over the fence kind of a Relationship between the AI and the human. And by that, what I mean is that it's a gig economy that anybody on Craigslist can sign up and see somebody's data and do this. So we wanted to really feel like this human in the loop was actually insanely high quality with a domain expert, maybe a tastemaker, maybe somebody like a CEO would hire so that even over time, as the AI learned from it, the AI actually got that much better and the human got to uplevel as well. And so it was insanely important for us not to have like a gig economy as set up. So we actually ended up interviewing a lot of these people before we hired them, contracted them to basically work with us with the AI, as well as building the tools that actually enable the agent to then delegate to the human and back and forth in a very secure, private way, so that the human only gets to see what they need to see in order to do that task. Um, so that was the first really hard setup to build end to end. And then the second thing, which I don't know if people even realize is, um, the, the age info, it's created to be very intelligent in that it won't just say, hey, now you do the whole task. It does what we call subtask delegation where it's if it's failing, say at a capture of a website, it will just escalate that and then take on the rest. And the humans actually as a result of that, are insanely efficient because now they don't have to take this whole mountain of a task. They just have to take something that they're really good at and the agent is not, and either decide to then train the agent for it or decide that actually this is a test making element that, hey, agent, you're probably going to learn, but not just by a couple of interactions. Um, so that was the whole design.
39:34Julian Weisser: There's some aspect here of, of, you know, trust that I think is really interesting in the world of AI, personal AI agents. Um, you know, I think that people always said, well, trust was going to become a really big issue with social media, and it turned out to really not be an issue at all. Most people were totally fine giving their data, giving lots of personal information. You know, there's some some people who who blur photos of their children when they post them on the internet. But even that, I think, is still the the vast minority. Um, so I'm really curious. Um, you know, if trust will actually become a much bigger issue now.
40:11Yeah.
40:11Julian Weisser: And my suspicion is it will be, um, because you have agents who make really embarrassing errors. And once that happens, um, the privacy and things like that become much more of a concern, uh, because they're less amorphous, like the fear. The fears are less amorphous. You actually have some agent do something that makes you look like a fool, and then you really, really don't like it. Um, so I'm curious how you think about privacy and trust. It seems like personal AI has a high degree of variability around this, and sort of all the people who are building. Some people thought that maybe their entire, uh, you know, like everything that they, they, you know, said to some of these, I won't say the name of the, uh, the personal AI, um, might be owned by the AI company. Um, I'm curious, uh, you know, to the extent, like what? What you think about this and kind of where you think things will that out is. I know that you really focus on privacy, security. What do you think is sort of the, the big the big opportunity here?
41:13Shivani Poddar: Yeah. I mean, it's, um, I think it's, uh, it's very interesting because, um, and I'll answer this in two ways. I think trust is actually multifaceted. One is, it is trust that, hey, my AI won't give my information to, uh, to somebody that it's just calling on my behalf and actually doesn't need to. I read that example that somebody's address was leaked on marketplace to someone they didn't want it to be leaked.
41:38They just showed up.
41:39Shivani Poddar: They just showed up. And so that's one aspect of trust, which I think is going to become incredibly important to people. And it takes, uh, it takes just one experience of having that happen to you to just like, get off that agent. This wasn't true in the social media age because there wasn't, to your point, any actual ramification of that happening to the, you know, bulk of the audience. Like, it was all theoretical, but in this case, it's experientially you'll lose trust with the society, with yourself. And so there's a lot of ramifications there. The other interesting part of trust is just that this will actually get done. And so I think that that so I don't think people choose any agent for this first one. I think people come to you for this second one. And man, it makes it so much nicer when the first one is insured.
42:31Yeah.
42:32Shivani Poddar: And so I do think that people choose us because they realize that, um, you know, it meets me where I am. It I could call it it can call it, but it really, truly gets things done. And. Isn't it nice that it won't give away my personal information to someone that I didn't want it to? And so I think that's how I think about trust, at least for now. But this is gonna we're I'm very excited to see where it evolves.
42:58Julian Weisser: Another way to describe it might be that, um, trust is a promise. And, uh, and a product or service is a promise as well. And if you live up to the promise, uh, then you increase the trust in the product or the service. Uh, if you don't live up to the the promise, uh, either by sort of falling short of actually delivering the outcome or you do something far worse, which is you send, you know, some stranger the person's, you know, personal residential address, uh, then you know, that that really, really tanks trust. Um, so so there are those two aspects of, of of trust that I think that really come into play here. And the first one is like, does it do what it says it will. And the second is, um, you know, does it does it also not only do that, uh, but also not sort of violate, you know, my privacy and sort of or make me feel like, you know, an idiot. Yeah. By doing something that makes me look bad. Yeah. Um, that's that's really cool. I'm really curious, you know, as we sort of, as we sort of head into the the tail end here of our conversation, it's been awesome. Um, I'd love to hear a little bit more about kind of how you've thought about, um, how you've thought about, uh, the dynamic of being a solo founder. You know, like, you, uh, you have an incredible team that's growing. I think you're hiring right now. So, in fact, you know, I know that browser based somebody's watching this, uh, the podcast with browser based discovered browser base ended up becoming one of the first hires on a certain certain section of their team there. Maybe somebody will. We'll be discovering fo through this. If they haven't seen the ultra viral, ultra viral launch. Um, and maybe, maybe they'll they'll be excited to join. Um, what are some things that you've learned about, you know, being a solo founder and building the team that we haven't covered yet, that that would be useful for folks both on the side who are starting companies, but also the folks who are looking at potentially evaluating joining them.
44:58Shivani Poddar: Yeah, I think as a solo founder, one of the things that's incredibly important is finding your
45:05Shivani Poddar: almost advisory bench. Or sometimes I say the the text, the people you text when you need help. Um, I think one of the things I realized, it's funny because when I was initially setting up my company, you know, all of us have to go through. Oh, who who do I, uh, who do how do I incorporate or how do I do? And I remember pinging somebody who actually wasn't a solo founder and they were like, oh, actually, like, this is something, you know, my co-founder would handle. And that's when it dawned on me that just the sheer number of
45:38Shivani Poddar: concurrent threads that solo founders have to carry is unprecedented. And so I ended up really, I mean, by virtue of some very good introductions from people, I ended up finding almost this set of speed text, like speed dials is how I associate that with who I'm like, what do you think? Who should I get? Or like, what was your experience with somebody? So I think solo founders have to augment themselves a lot. Um, and I think that can be either with people who've invested in you that can be by communities. Um, that can be just by finding founder friends that you've just come across and ended up building a good relationship with. I have one who, uh, she and I haven't ever really met, but we, like, keep texting about our experience, and it's been the most helpful relationship through building this company because she's done it as a solo founder, too. And so I think, um, that is the singular most important advice is thinking about how do you augment the things that you're not. And I think that also then sort of translates to hiring in that I
46:46Shivani Poddar: one of the things I've realized is every solo founder has like a strength and maybe it's technical, maybe it's, um, marketing, whatever that is. It's incredibly important to figure out who that yin to your yang is, and then find the best ever person who can carry some of those things. And this won't be one person, by the way, because otherwise you'd have a co-founder. But this would be a team of those people who can then complement you in the right way.
47:13Julian Weisser: How do you think about that on your end? So when you're building sort of the first five, what are you hiring? The sort of the first five roles onto your team? Um, were you thinking more opportunistically about these are the people that I would want to have on my team in any capacity. Uh, was it more like. I think there are these very specific types of things that, you know, speaking to your strengths and also deficits because everybody has them. What were the things from a practical standpoint, if people were, you know, you know, listening to this, they wanted sort of a case study in your case. What was that? And how did how did you think about those roles?
47:45Shivani Poddar: I mean, I knew that I was building something that was going to be the biggest technological challenge that existed. Um, and I, I was actually starting from a great place because I've built technologies before. At the same time, I knew that I am not the one who's going to be building this insanity by myself. And so for me, it was definitely starting with where the core of the business was going to be, which was technology. And so hiring the best people who could do parts of that stack really well. Um, and then obviously I think, uh, when I think about hiring and at some point I will share more, but it's about, I think about what do we need that's going to make us insanely successful, not just in the next three months, but which is going to be our unfair advantage in the next one year. And how can I work backwards and hire those people for that unfair advantage that we're going to create in a year from now? So I think about hiring a lot from not just the now, but the future. Um, and again, a lot of these bets in that way pay off. So projecting out your product, your business, your company ahead in the future and then hiring now for that future sometimes is also a very interesting way to hire which which can feel very counterintuitive because as a solo founder, you feel like the ship is burning and I need to put out the fire all the time, but sometimes it's just so that you can think about what's the next shore that I need to build for.
49:12Julian Weisser: And I love that you talk about having a larger equity pool for early employees, because it's so important to do, especially when you consider the fact that when you don't have co-founders, there's a lot more responsibilities spread out across a wider group of people. Um, I'm really curious. Um, you know, if you were to sort of evaluate kind of how you spend your time as a solo founder, kind of on a percentage basis or something like that, how has that changed from kind of early days to kind of where you are today?
49:41Shivani Poddar: Yeah, I mean, that's such a I've been thinking about this a lot because I think about spending time and focus a lot. And maybe this answer is sort of what's expected is the early days are just about hunkering down, building the product, um, bringing this technology into the world. I actually remember the first time that our agent actually made a call to order something from a restaurant. It blew our minds away. And I remember all of us were giddy together, and we were actually, like, on a live call. And we could not believe ourselves. And the fact that sushi showed up.
50:17Shivani Poddar: And so obviously the first few months is just everybody sort of breathing the same vision and product and focus. And now, over time, as the product has taken on a life of its own, it's, um, for us, it's, uh, for me, it's about how do I bring that product to the world? I spend a lot more time on those parts of the mission. How do I now hire more good people? I spend, uh, almost as a way of headcount. I have to spend way more time hiring way more people. And I think, um, spending time to uplevel, uh, my team in a way where we all are sort of, you know, doing more than we were, in a way. Um, so, yeah, that's that's kind of how the journey has evolved.
51:04Julian Weisser: This has been awesome. Uh, I cannot wait for more people to be using what you're building. Uh, it's truly remarkable to see, um, sort of the confluence of all of your past experiences, um, you know, the new technology and what it unlocks. And then also this idea of bringing in humans to, uh, to both have the I complement them and for them to complement the AI, it creates what looks it feels like one of the the best in class sort of experiences that could really help people, uh, save time on things that really matter so they can spend time on things that really matter. Um, we close with two customary questions. Uh, the first question is around sort of the case against solo founding the bear case for solo founding. Well, I'll posit that you, uh, you you share sort of the bear case for it. Uh, we'll talk a little bit about that and then prepare you after that. We'll give the case for solo founding.
52:04Oh, wow. So.
52:06Julian Weisser: So what's the reason? Uh, maybe somebody should be hesitant about being a solo founder, uh, at this moment in time.
52:12I mean.
52:12Shivani Poddar: I think it's just founding is hard. There's troughs and valleys and, um,
52:19Shivani Poddar: even if it's funny because, you know, when when our product hadn't launched yet, we were like, oh my God, we wish the launch goes well. Oh my God, we wish we have like hundreds of thousands of users. Like that's all you can think about. And that's like the hard part is like building up to it. And then it's funny you launch and it goes really well, and then you're like, oh my God, there's so many users. We need to get more compute. We need to like scale more. And so it never really gets easier. The founding journey never really gets easier. And so I think, uh, it can get uh, you definitely, as a solo founder, have unique stresses that you have to carry alone, um, in a way. And so I think the bear case of a solo founder is definitely, uh, the ability to sort of be an island that can self-regulate and self, um, anticipate the future and all of that comes with it.
53:14Julian Weisser: Do you feel like that's something you have developed over the months or learned about, or do you think that that's something that is kind of sort of stylistically who you are and and you sort of default to that? Like, do you feel like you were self-regulating when you were at these larger companies or. I'm curious because I know there are some things, for instance, not, you know, if you say that sort of gaining access to resources is something that founders need to be really good at. Yeah. Um, you know, accessing capital, accessing, you know, getting in front of customers. Right. Recruiting people to join your team. Um, some founders could be excellent at all. Three on day one. Some of them, you know, are much better, much spicier on certain ones that really matter but need to, you know, need to hone in and get better on others potentially. So I'm curious about that, about that aspect for you. Was that something that you've learned? Is there anything that you've, you know, you could maybe impart to folks about that, uh, and what you've experienced with it so far.
54:13Shivani Poddar: I mean, I've had to learn on all accounts because it's funny. I used to joke. I was, for example, I, I used to, uh, joke that I can hire anybody in the world. Um, and I think that's that's largely true, but not true when you're actually hiring for a no name startup, especially when you're in stealth. Like, nobody knows what you're doing. Nobody has, like, quite the idea. And so I realized that I had to, uh, really work hard to transfer those skills to the startup world from the big tech world. And so, um, there was definitely lots of learning. I do think that there's still a lot of learning. I do think that I've had to build these muscles, and I've also had to lean on support systems around me. Um, if my partner was up here, he would be like, hell yeah.
55:01Shivani Poddar: What do you mean? And so I do think that, you know, our loved ones who see us go through these journeys, like, carry on a lot for us a little bit, and then the rest of it, like we learn and we evolve. And that's partly one of the journey as you start somewhere. And then when you're on the other side, you're like, oh my God. Like, look how far we came.
55:24Julian Weisser: That's amazing. Okay, I guess, I guess, like, the last thing to take us home is the case for solo founding. If you if you were sort of talking to a solo founder or prospective solo founder right now, somebody who's thinking about doing it, um, you know, they have the right motivations. But what would the case for being solo be?
55:43Yeah.
55:43Shivani Poddar: I mean, I think, um, I, I, I had to do a lot of this thinking. I have come to realize, and I've actually spoken to a lot when I was actually going through this, I spoke to a lot of people who actually didn't start solo, but ended up having a really bad found co-founder breakup and ended up being solo. And what that journey told me is I'd never be against having a co-founder, but I'd never rush into it just because I wouldn't want to do it solo. And so I think the cost of actually not making that dream or that idea come true, is way higher than the cost of learning how to self-regulate or building these other muscles. Especially now, because I do think that what the AI tools of the last few months even have afforded us is genuinely an opportunity that has not existed ever before. And I know solo founders who were very technical, but now can build the best presentations for their pitch decks, or founders who were actually not technical at all, but are able to build prototypes to attract the best engineers. And so the the very, very bulk ish for solo founders is just go do it. And I think if if any of us are good solo founders will figure out how to find AI to like, do the rest with us.
57:06Julian Weisser: What a great way to end it. Don't defer your dreams. Thank you so much for being doing this and being on this.
57:11Shivani Poddar: Thank you for having me. This is such a good conversation.
57:15If you enjoyed this conversation, we would love it. If you can share solo founders with everyone you know, particularly people who are considering starting companies solo or who are already doing it, you know, the entire goal of what we do is to normalize solo founder and encourage more people to solo founded instead of end up with co-founders of convenience, if you'd like, and we'd really appreciate it, you can go. Leave a review on Apple. You can go give us a thumbs up. You can subscribe on the YouTube channel. Anything helps. And of course, if you are starting a company or considering starting a company, you should go look at the Solo Founders Program. We work together with you in San Francisco alongside of a bunch of other solo founders. The idea is it's much better to be solo together than to be solo alone. I'll spend a lot of time helping you, and we also invest $100,000 in your company. So if you're excited to learn more about that, go check out Solo Founders com Slash program.


