Aravind Srinivas: The Race to Build the AI Browser of the Future

Aravind Srinivas on June 16, 2025 at AI Startup School in San Francisco. Aravind Srinivas started Perplexity with one goal: to rethink how we search, browse, and interact with information online. In this conversation, he shares the journey from hacking together a natural-language-to-SQL search tool to building a product used by millions around the world. He talks about the big bet on the AI-powered browser, why agents—not just chatbots—are the next step, and how speed, accuracy, and focus help a startup compete against giants. Along the way, he reflects on co-founder dynamics, early technical challenges, and what it takes to keep building when the biggest players in the world are racing alongside you.

• 0:00Speaker 00

  YC's next batch is now taking applications. Got a startup in you? Apply at ycombinator.com/apply. It's never too early and filling out the app will level up your idea. Now on to the video.

• 0:13Speaker 01

  You have to innovate, you have to move faster than everybody else. And it's like running a marathon, but at an extremely high velocity, right? The only mode you have is speed. I read all the Twitter comments every time. Google I/O last year was AI overview and perplexity is dead. This year was AI mode and perplexity is dead. And I read all of that too and it's always fun. I love it actually.

• 0:39Speaker 02

  Ervin, I see you every, I don't know, two or three months and you give me an update on the latest on perplexity. Why don't you just tell these folks where you're at? How are things going? Do people use perplexity? Do you guys use perplexity?

• 0:55Speaker 01

  Well, whether you believe it or not, like I have infra issues every day. So there are a lot of people using it and this usage is actually growing. To the extent that we don't actually know how to deal with it. We have to rebuild the infra to scale the next 10x. So definitely a lot of people in the world using it, thanks to all of you as well. What is next for us? The browser, that's the big bet we are making as far as the future of the company goes. Everyone's here is like, why should I use perplexity when there's search and other AI apps? Of course, ChatGPT has a bigger distribution than us. Every other AI app is trying to put search as a layer in it. All of them support citations. A lot of them support some of the verticals we have put work into.

• 1:40Speaker 01

  Yes, like we're always going to continue to remain better than others in that category. But I think the browser and agents are truly the next bet that we want to make. We think about it as an assistant rather than a complete autonomous agent. But one omnibox where you can navigate, you can ask informational queries and you can give agentic tasks and your AI with you on your new tab page and your sidecar as an assistant on any web page you are. Makes the browser feel like more like a cognitive operating system rather than just yet another browser. And we hope to make it like a cloud where you launch several tasks in parallel that are running asynchronously. And pulling all your personal contacts,

• 2:25Speaker 01

  your email, your calendar, your Amazon, your, you know, all sorts of social media accounts that you have. And you go and do your research on real estate, the markets. And these are all like just processes running on your browser. That's never been possible before. And Chrome was exciting when each tab was its own process. You think about each query or each prompt could be that. And that will be our new browser comment. So we're putting all our energy for that.

• 2:53Speaker 02

  This was gonna be the hard question I saved for the end, but since you queued it up, I'll do it right now. I think if Sam Altman were still on the stage today, he would say, oh yeah, that's what we're doing. And I think Sundar at Google probably would say that's the direction we're headed as well. So it feels like there are a bunch of players, now many of them very well-funded, going in generally the same direction. How do you see the world? Like, do you think that there's gonna, it's gonna play out where there's actually like a bunch of different use cases? And you can own a very important one that others won't want to own? Or are we in for like a major competitive battle?

• 3:31Speaker 01

  Look, if something is really worth doing, it's really natural that people with a lot of funding will go and do it. People said, Perplexity is a great product. Now everyone is trying to do something that can answer any question with sources. Cursor was a great product. Now OpenAI is trying to buy Cursor's competitor, Anthropic Launch Codex, like Claude Code. Google has its own like rival tools. So it's only natural that when there's a lot of money to be made in a certain sector, people are gonna try to copy it. And there's only a limited amount of things you can be world-class at. Whether it's being, building great models or building one or two really good products. So you're obviously not gonna win on everything.

• 4:15Speaker 01

  For us, this is the only thing we care about. Accuracy at the level of answers, accuracy at the level of tasks, orchestrating all these different tools. The browser is much harder to copy than like yet another chat tool. That said, I'm fully... Working with the assumption that OpenAI will also build its own browser, Anthropic will also try to build its own browser. Google already has one called Chrome, so it's completely reasonable to expect them to do it. And the only mode you have is speed. You have to innovate, you have to move faster than everybody else. And it's like running a marathon, but at an extremely high velocity, right?

• 4:53Speaker 02

  Yeah, I really agree with your statement that like, you can only focus on one thing and be world-class at one thing. And just to give you guys like a little glimpse into it, we were backstage before this talk and he was showing me some of the new stuff that they're working on. And there was like a bug, right? And he stopped everything he was doing to like figure out what was wrong with this bug. Why was it not doing the right thing? And if you think about like, what would the CEO of a large company do in that situation? Probably they would like hand it off to somebody else on their team. So it's like a good piece of evidence that you actually mean what you say.

• 5:26Speaker 01

  Yeah, I love triaging and fixing bugs. I know it sounds trivial, like is that the best use of the time of CEO? There are a lot of people who would think otherwise. Recently, people are like, oh, like they're, I hope this behavior is rubbing off on others. Like I've noticed even Sundar is doing bug support on X right now. So I'm happy that like, you know, that's setting a good example.

• 5:49Speaker 02

  Right. Okay, let's go back to the beginning. Like, most of the folks in the audience here are either students or recent grads or grad students. And I think hearing your story of like how you started perplexity would be really interesting to them because it's probably exactly the world that they're in now.

• 6:05Speaker 01

  Yeah.

• 6:06Speaker 02

  Tell us how you got started.

• 6:07Speaker 01

  We started the company without actually having a clear idea of what to build, which is the opposite of what YC advises, which is start from a project and turn it into a company. I really think at this point in time, when AI is improving so fast, you don't have to rigidly stick to any one idea when you're getting started. But the most important thing is you don't change the idea every week. Like that you shouldn't do either. So. Start with something like brainstorm, think about it and then try to immediately build it and get it in the hands of people. Uh, one tool that we were building was natural language SQL. Which we actually thought about it as a search tool, searching over relational databases.

• 6:50Speaker 01

  I loved Twitter search, but never like, like the original version of Facebook graph search. I loved that when I was much younger. So I wanted to like rebuild that by using language models. Um, and I love Twitter as a platform, so it was, there's no good way to search over Twitter. There still is no good way to search over Twitter. But at least at the time, we organized the entire Twitter's data in the form of like relational tables. And just converted every user's query into a SQL query. And ran it as against the database and was really, really good. And um, that's what got us started. But at some point, we figured it's better to like scale this across the web.

• 7:32Speaker 01

  And we cannot make every website in the form of tables. And neither is it actually easy to answer all sorts of questions. So we bet on the fact that language models can do all the reasoning and parsing and like structuring later. But the more important thing is to start with something more unstructured. And that ended up becoming perplexity.

• 7:52Speaker 02

  Got it. Um, and maybe one step before you actually left to go start the company. Like, how did you find your co-founders? How did you decide that? Machine learning and AI was like the area you wanted to focus on?

• 8:05Speaker 01

  Cause that's the only thing I was good at. I was not good at anything else.

• 8:08Speaker 02

  Okay.

• 8:08Speaker 01

  So what's the point in starting company? I cannot start a delivery company or a social media company. Like I'm not, I'm not the right fit, right? Uh, the only thing I knew was AI and machine learning. In fact, it's funny, I, we started an AI company. Uh, but we not made fun of like not even training our own models, like more, but only the foundation models or stuff we don't train. We trained so many different models. But uh, that's the extent to which you need to have the intellectual humility to know like what you're good at. What is actually doable for you within, with the resources that you have access to? And the co-founders are like people I, like, like I knew from grad school. So we'd been talking and discussing ideas for a long time and um, I think grad school is a great way to like, uh.

• 8:53Speaker 01

  You know, like, like identify your co-founders. You don't talk to them with the, you know, with the long-term calculation of like, oh, this could be my co-founder of my future company. You talk to them because they're interesting people. And I think that's essentially the value of the Y Combinator network. So even if your first startup year fails, you get access to a lot of amazing people and maybe they could be a future co-founder. So that that's essentially what grad school was for me.

• 9:17Speaker 02

  Yeah, that's awesome. Um, okay, so you launched this first version of perplexity, which is largely to like do Twitter search effectively.

• 9:25Speaker 01

  Yeah.

• 9:26Speaker 02

  At what point did it like start to work? And you maybe internally felt like, oh, we should keep working on this. This is gonna be something to explore.

• 9:35Speaker 01

  Yeah, so whoever we gave early access to, they were all very excited about it. They kept using it repeatedly. I think there is a phenomenon in products where there's an initial wow factor.

• 9:45Speaker 02

  Yep.

• 9:46Speaker 01

  And then mostly either drops completely, that that means you never had real retention. Or it it definitely drops, but there's sustained usage. So when we saw that for the relational database searches like Twitter, LinkedIn, GitHub. We knew that we had, like, there was something magical about combining large language models in search. But then what we did is like, we dreamt bigger and said, what if we just give answers and cite the relevant sources? We launched that as a Discord bot and that was also continually being used. It was not like a one-day usage and people started ignoring it. So that's when we decided we had the courage to launch it. We launched it seven days after the ChatGPT launch. Especially at a time when ChatGPT did not have web, right? Web search, that was a good moment.

• 10:30Speaker 01

  And I think like many of the successful AI products that people speak about today, cursor included, all were like twenty twenty-two launches. Or like, like early twenty twenty-three or late twenty twenty-two launches. So they're all like old people, you know, in the, in the, in the, in this AI time scale.

• 10:46Speaker 02

  Yes.

• 10:46Speaker 01

  For me, the aha moment was like the New Year Eve, there was like close to seven hundred thousand queries. And I was like, okay, this has the crappiest name for a consumer product. It's called perplexity, very hard, nobody even knows how to share it. And then it was so slow, took seven seconds to answer for a query at the time. And um, it was making a lot of mistakes, hallucinations. And like a no name company, no name founder, very one or two million dollars in seed funding. Despite that, people were caring enough to sharing screenshots and like, and then on New Year's Eve, when you could be, you know, watching Netflix, right? So that's when I knew there was something real here and I started like optimizing for like, you know, committing to this this mission.

• 11:30Speaker 02

  Okay. And at that point, like on that New Year's Eve, did you in your head think? I'm building a thing that could really compete with Google and like take over a market as big as what Google offers?

• 11:42Speaker 01

  No.

• 11:42Speaker 02

  Or was it just a toy for you?

• 11:44Speaker 01

  The first time the thought occurred to me was when Google wrote a blog post, Sundar wrote a blog post about Bard. Like that was around the time when we were raising series A funding. And everybody said, okay, Bard is gonna do whatever you're doing. And I said, why do I have to build Bard? Like, why not just do it on Google? Right? Where you have all the distribution in the world. So why do you have to build a separate product? Like just just update your core main?

• 12:11Speaker 02

  Yeah, you have the best possible asset to do this.

• 12:13Speaker 01

  Exactly. And when I kept thinking, it was pretty obvious. You cannot like if if if people can get answers to? Best hotels to stay in San Francisco with a view of the Golden Gate Bridge or like Bay Bridge? Or like where can I stay in New York, like next to the Central Park with good amenities? Or like which flight is the best thing for me to take to fly from SF to London? If you get direct answers to these questions with booking links right there? How are you gonna mint money from booking and Expedia and Kayak and like? Like, you know, all like same same same thing for shopping? How are you gonna take money from Amazon and like Walmart for the same ad where they're all bidding against each other? It's not in their incentive to give you good answers at all.

• 12:58Speaker 01

  So that's when I realized that they have to build a separate product, but they can never capitalize on their core distribution. And twenty twenty four, twenty twenty three especially, and a large part of twenty twenty four too. Google had like maybe the fourth or fifth best models at any moment. So as a startup outside Google, you had access to AI that was better than what Google internally had, which was unprecedented, right? Until then, if you had to compete with Google and you had to build something that needed a lot of AI in it, good luck, right? Like, because you never have an AI outside Google that's even equal, leave alone being better. But now it's a complete reversal of the situation, thanks to OpenAI or Anthropic or open source models.

• 13:44Speaker 01

  So that plus innovator dilemma, plus the fact that we could make a lot of mistakes and it's fine. Whereas for Google, one mistake tanks their stock. Like, you remember the live demo of Bard where it failed and the stock went down by 6%?

• 13:57Speaker 02

  Yep.

• 13:57Speaker 01

  So we knew that there was a lot of advantages for us.

• 14:00Speaker 02

  Yeah, I heard you talk about this recently, but, you know, Google specifically has been trying to build perplexity-like experiences. And, you know, you joke-

• 14:10Speaker 01

  AI mode?

• 14:11Speaker 02

  Yeah, they just like changed the name of it each Google I/O.

• 14:14Speaker 01

  Yeah.

• 14:14Speaker 02

  And then not really launching-

• 14:15Speaker 01

  It's true. I'm not, I'm not, I'm not like saying something wrong.

• 14:17Speaker 02

  No, it is. Yes.

• 14:18Speaker 01

  Right. So it's like, I, I, look, it might sound a little cocky to say that, but it's true. The same feature is being launched year after year after year with a different name, with a different VP, with a different group of people, but it's the same thing. Except maybe it's getting better, but it's never getting launched to everybody.

• 14:39Speaker 02

  One of the things I've come to admire about you is you really have a focus on the user experience. And, and you told me how you kind of learned that from Larry Page.

• 14:49Speaker 01

  Yeah.

• 14:50Speaker 02

  By reading the book about Google. Why do you think Google has lost that ability?

• 14:58Speaker 01

  Well, it's a much bigger business, right? And it's not founder-led anymore. It's hard to take risks. I think they have great people. Nobody like, no one in this audience would think Google has incompetent people. I think they're like really great engineers. It's largely the incentive structure. It's hard to like, you know, take a hit on your own stock and do the thing that's long-term correct. So, you know, honestly, I'm happy that that sort of dilemma exists because otherwise, where is the opening for startups, right? And then if startups can succeed, then it's gonna be monopolies getting bigger and bigger. And that's not great for the world. I actually am very happy that we are able to win and they're also like able to like ship new products and people are like first time comparing, right?

• 15:43Speaker 01

  Earlier, for access to information, you would never even bother to compare an alternative to Google.

• 15:49Speaker 02

  That's true.

• 15:50Speaker 01

  Like that was like considered a waste of time, a joke. Now, at least you're like, oh, I first go ask this app, like I'll ask Google, or I'll ask ChatGPT, or I'll ask Perplexity, or I'll ask Gemini. And then maybe you don't even ask Google anymore, you just ask the AI apps. And there are a bunch of AI assistants and the phone makers will start offering all of them as alternatives. It's not gonna be like a locked in default search option. So I'm really happy that they're competing in a world where a monopoly hopefully doesn't exist. And that creates a more fair ground for everybody.

• 16:23Speaker 02

  Yeah, yeah. Yeah, you were also telling me backstage about, you know, you are facing this increased competition from a variety of folks. But if you look at your numbers, you haven't really seen the effect.

• 16:34Speaker 01

  You know, I read all the Twitter comments every time. The Google I/O, exactly the same set of comments repeated this year. Google I/O last year was AI overview and perplexity is dead. This year was AI mode and perplexity is dead. And I read all of that too, and it's always fun. I love it, actually. Because like, like, they know that, they're all thinking like, I don't, I don't even expect these things. Or the people in the company are like thinking Google wouldn't build this or something like that. But it's, the reality is like, nobody actually gets exposed to those features. But competition is real, okay? Let's assume, let's, let's, let's. Except that OpenAI is extremely well-funded, doesn't have all these innovative dilemma problems, wants to actually ship search on the ChatGPT.

• 17:19Speaker 01

  ChatGPT is the most successful consumer AI product out there. And so, competing against it is very difficult. Which is why I, I really wanna like, uh, push the company more on the browser side. And I think Comet, the browser, will be an abstraction layer above chatbots. I, you could even imagine like, I, you know, if you permit Comet, all your ChatGPT chats can, you know, be fed into that AI and like, you don't even have to worry about memory or personalization or like, you know, any of these things. And it'll do a thing, a lot of new things that a chatbot cannot do. Like accessing other tabs, accessing your browsing history, going and completing forms for you, like paying your credit cards, buying stuff for you.

• 18:03Speaker 01

  Uh, and being your scout, you know, going and doing all the research for you. That sort of thing, like periodic recurring tasks. I think that's the magic that the browser enables for you. And putting it into like mobile, like, like building mobile versions of this browser is gonna be very hard, like just engineering-wise, it's gonna take many months. So I'm not really worried about like someone else trying to copy this, it's gonna take time for anybody.

• 18:26Speaker 02

  Switching to a different browser is like a pretty big decision for a user. What, what do you think will be the very short-term things that your browser will do so much better than what I can get today in Chrome that will make me wanna switch?

• 18:40Speaker 01

  The perfect blend of AI, navigation and agents is, is, is what we're gonna offer. And, um, might sound like a boring answer, but no one's done that. And there are like hundreds of millions, probably close to a billion people using AI these days. So the market's already pretty big.

• 18:57Speaker 02

  What's like a specific example of how I would do that, you know, if I had access to it tomorrow?

• 19:02Speaker 01

  You can schedule your meetings. You can reply to some of your emails that you don't even want to read. You can, like, for example, let's say you're hosting a Y Combinator event and you say, I only want to accept Stanford dropouts. And it can go through the entire list of people who applied and just filter based on who's, in order to scrape their LinkedIn URLs.

• 19:22Speaker 02

  Yeah.

• 19:23Speaker 01

  Filter based on whether they were Stanford and whether they dropped out or not. And then accept it, like that level of multi-step reasoning.

• 19:30Speaker 02

  Got it.

• 19:30Speaker 01

  Is something you can uniquely do. By the way, I'm not saying that's a good filter. I wouldn't get it otherwise. So hopefully you're more open, yeah.

• 19:39Speaker 02

  We look for deep-minded researchers also.

• 19:41Speaker 01

  Yeah.

• 19:41Speaker 02

  Yeah, don't worry. Okay, cool. Let's talk a little bit about how you run the company now, right? I don't know if you wanted to say how many employees you have?

• 19:51Speaker 01

  Yeah, we have about 200.

• 19:52Speaker 02

  Okay, so the company's getting bigger. You now have access to code-writing AI tools. Are you guys just like full in on that stuff? Are you vibe coding everything? What's it look like?

• 20:04Speaker 01

  I mean, you don't want to vibe code everything, right? Like, we frequently run into infra issues and you don't want a vibe coder right there fixing it on live things on production. Like, I do want like people well-trained in regular software engineering, infrastructure, distributed systems. Like, you don't want to replace these skills. But yeah, front-end design, that's where we're seeing tremendous adoption, like cursors being used by everybody. We made it mandatory to use at least one AI coding tool and internally at perplexity it happens to be cursor and like a mix of cursor and GitHub Copilot. But yeah, we definitely made it compulsory. And so the way machine learning people are doing, using it, AI people are like, sometimes

• 20:48Speaker 01

  they read a paper and they can just upload a screenshot of the pseudo-code. And ask Cursor to like just edit the files to implement this new algorithm. And then it's able to like write its own unit tests and then run an experiment pretty quickly. That is reducing experimentation time from like three, four days to like literally one hour. Or like there are people who don't know design and so sometimes I just give them feedback. Where I take a screenshot of my iOS app. And I said, this button needs to move here with an arrow and they upload my screenshot to Cursor and then ask it to like write, write a change to the Swift UI file.

• 21:27Speaker 01

  So that level of change is incredible, like the speed at which you can fix bugs and ship to production is crazy.

• 21:36Speaker 02

  The more bugs there are, as long as you can fix them fast.

• 21:38Speaker 01

  Yeah, bugs are always ahead of like how fast people can write code though. But just to be clear, I'm a big fan of all these tools. But it is also introducing new bugs and many people don't know how to fix them. And they don't even know how the bug got introduced and they have to go find it again, so it's not perfect. And I actually like the more newer tools like Claude code seems to be far smarter than like what Cursor is able to do. So I'm actually like, like really positive that this is the right future, but there are, there are issues right now.

• 22:12Speaker 02

  Yeah, in talking to a lot of folks here, one of the major questions that I've heard is as these coding tools get better and better and better. What is the like actual enduring value of a company like yours? If, if increasingly it's easy to replicate what you have done using these tools. What's your take on that general type of question?

• 22:35Speaker 01

  I mean, brand definitely has a big value, right? Like there are Cursor competitors, perplexity competitors, like OpenAI will have like their own Cursor. OpenAI has perplexity within ChatGPT that did not kill any of these companies. So there is a certain brand value that once you acquire at the scale of like several millions of users, paying users. You don't actually die that fast. You earn the right to survive and keep building. So brand is important. Narrative is very important to the brand. Like you have to communicate to people, why do you even need to exist? For us, it's the focus on accuracy. Okay, there exists hundred chatbots, but we are the most focused on getting as many answers right as possible.

• 23:20Speaker 01

  We focus on speed, time to first token on app or web, like we're still the fastest despite doing search. Uh, we focus a lot on like how we present the answer. So there are some things you are obsessed about because you care about it. And that becomes your narrative and your brand identity. And if you manage to get reasonable amount of distribution, not saying hundred million users, but tens of millions. Then you earn the right to keep playing the game, no matter what other people ship. Until then, it's definitely a challenge. You have to worry about it. Even now, we worry about it and the only solution is to move fast and keep shipping.

• 23:56Speaker 02

  Beyond brand, like, do you think about any network effect types of things emerging with perplexity?

• 24:04Speaker 01

  And brand has network effects, right?

• 24:05Speaker 02

  Sure.

• 24:05Speaker 01

  Like people, people tell each other about the brand. But no AI product has within app network effect. Like, like, it's not like WhatsApp where if you build a WhatsApp rival. Meta has a definitely like a questionable brand, right? Like people don't necessarily trust Meta's products. They think like these are ad products.

• 24:25Speaker 02

  Hmm.

• 24:25Speaker 01

  Despite that, nobody is able to switch off WhatsApp that easily. Because there's all your contacts, your groups, everything's there. AI doesn't quite have that yet. Uh, mainly because you can easily export your chat GPT history, upload it somewhere else or uh, things like that. I think the browser will definitely be one play to like. You know, figure this out because as you're browsing history and like, which again, you can still export, but not the same as just getting a dump, a CSV dump. Uh, and your passwords, your wallet, your agent remembers you, there's a lot of tasks that are running on the browser. That you rely on your day to day life and work. That's one way to like keep get the product a lot more sticky and like create more network effects.

• 25:09Speaker 01

  Especially if multiple people rely on the same set of tasks, you're sharing it with them. That's one way to get all this into like in the next level.

• 25:16Speaker 02

  It it also sounds like a lot of the stuff that you aspire to solve for users requires integrations or partnerships or something with a bunch of other companies in the world.

• 25:26Speaker 01

  Yeah.

• 25:26Speaker 02

  And if you can get those to be good, then there is somewhat of a network effect in the sense that your product will be good and some competitor would have to build the same integration or same deal with with these providers. What does that look like, do you think, in the future? Like, does Perplexity do deals with all the airlines in the world and all the hotels and all the e-commerce providers?

• 25:49Speaker 01

  So we already work with Selfbook. They power all the hotel bookings, natively done on Perplexity. We work with TripAdvisor to surface all the reviews of hotels and different, you know, places. We have like collaborations for the maps. We work with Yelp. We also like, you know, for shopping, we have a lot of merchants who are directly selling on us. And then we work with Firmly to like support the bookings, like native purchases. So there's already a lot of partnerships. Shopify is one of our partners. On finance, we work with FMP. Sports, we work with like Stats Perform. So there's a lot of data providers already working with us to on these verticals and we just think it's going to expand further

• 26:34Speaker 01

  as agents start to do things. Different people are okay with like becoming MCP servers. Some people are not. Some people just want to like like preserve their websites. The browser agent will be generic enough that it'll respect whatever the third party wants. Because at the end of the day, the agent is the one that's being permitted by the user to act on their behalf. And if there is no MCP server, it's still fine. You can just use these tabs as if the user would have done it. And that's the key advantage of the browser that you do not have if you commit entirely to just MCP vision. If you commit entirely to MCP vision, you require these third-party MCP servers to work reliably.

• 27:17Speaker 01

  The data that they send you on with the MCP protocol has to be perfect. Your chatbot has to like, like, you know, deal with all these issues that exist. On the other hand, if you just ground up, design it as the way a human would use that website, you have full control over like how to, how to do it. You don't have to rely on someone else doing engineering well on their end.

• 27:40Speaker 02

  Let's talk next about business model. Your main competitor, Google, their business model, as you've talked about, is selling ads. And you think that prevents them from being really good at what you're doing. So what, what will your business model be and how will you get it to be on the order of magnitude of, of Google's?

• 27:57Speaker 01

  I don't know if we'll ever get order of magnitude profits as Google, just to be clear. And I don't think that's needed. No one in the history, even Google themselves, never has had another business that had the margins that Google has. So it's completely reasonable to get something far, far better than any public company out there right now. And, and be still way below Google. Number two, I think the subscription revenue is like really encouraging. We never expected to get this far. And then we think like we can grow at least, you know, a few billions a year in just subs, which is a great business. Usage-based pricing, where people are paying an agent for completing a task.

• 28:39Speaker 01

  Or people have recurring tasks and they pay based on, you know, every single use of the task. And they normalize this system based on like how much it would take to hire a person to do that for them. It's going to be a thing. I don't exactly know how it's going to play out, but what are the margins going to be on that? Potentially, it's going to be way better than subscriptions in terms of volume of people who would pay for it. But it might be lower margins because it's usage-based, so you're still going to be spending on all those queries. Someone might be paying a subscription to like one of these AI apps and might not have used it for an entire month. And so that's good margins on that user, right? So I don't actually have a clear sense of how this is all going to like evolve. But all I know is subscriptions and usage-based pricing are going to be a thing.

• 29:24Speaker 01

  Transactions, you know, if people start buying more through AIs, taking a cut out of the transactions is good. It's going to be CPAs have historically been way lower margins than CPCs, which is why Google never became a transaction platform. Which is why I said, like, you're gonna make a lot of money here. You may never make as much money as Google.

• 29:45Speaker 02

  Yeah, Google, Google's business model is potentially the best business model ever.

• 29:49Speaker 01

  Ever, yeah.

• 29:50Speaker 02

  So, yeah, it's fair to not.

• 29:51Speaker 01

  Maybe it was so good that you needed AI to kill it, basically.

• 29:54Speaker 02

  All right, let's, we're gonna do some audience Q and A in a little bit, but before we get there, I kinda wanted to understand your advice for the folks in this room, right? Like, if you were in their position back, whatever it was, four years ago, what would you advise they do?

• 30:10Speaker 01

  I would say, uh, work incredibly hard. There is no substitute for it. Don't think like you're very smart, like, like strategizing the right way to build a company despite all, like, what big model labs are doing. You should assume that if you have a big hit, if your company is something that can make revenue on the scale of hundreds of millions of dollars, or potentially billions of dollars, you should always assume that a model company will copy it. Mainly because they are really looking for revenue. They've raised like tens of billions or close to fifty billion, and they need to justify all that capex spend, and they need to keep searching for new ways to make money. So they will copy anything that's good.

• 30:53Speaker 01

  I think you gotta live with that fear, you have to embrace it, and realize that, like, your mode comes from moving fast and building your own identity around what you're doing. Because users at the end care, like, when you're trying to get, like, a specific person for your house help, you are searching for that specific person. You're not, like, going for a general agency that handles all of it. So, uh, I think there's, like, like, real benefit from embracing that fear, and, like, sleeping with that fear, and waking up every day and, like, feeling excited about what you're gonna build, because that's the only thing that'll keep you going.

• 31:27Speaker 02

  Well, you guys are the perfect example of how it's possible to go up against somebody as big as Google. That's great. All right, let's do some Q and A. We'll start on the left side here. Go for it.

• 31:38Speaker 03

  Hi, my name is Sammy and I just want to personally thank you for helping me get a 100 in my Theory of Knowledge course. Would not have been able to do it without you, no shame. Quick question for you. You know, with your recent partnership with Nvidia to ship AI models across Europe, there's been talks about perplexity being installed on all Samsung phones or pre-installed. And that could lift your valuation towards 14 billion, according to sources like Bloomberg. It's a heavy responsibility, being the default search engine, for, you know, the mainstream population. What do you think are the most important factors at perplexity to prevent hallucinations or incorrect data from, you know, being given to the masses? Thank you so much.

• 32:23Speaker 01

  Thank you. Hallucinations is something we really care about. We're building benchmarks internally to keep up to date with that. The only way there is to keep building a better search index, keep capturing better snippets of all the webpages, and then like, these models like, you know, are getting fast enough that you can have them reason multi-step for every query without incurring too much cost. And so that's another way to reduce hallucinations.

• 32:50Speaker 04

  I want to ask you about like, the innovator's dilemma. So, if you were in Sundar's shoes, or in, like, the Google co-founder's shoes, like, what would you do? And how would you kind of come up with maybe changing their business model, even if it's a worse model?

• 33:06Speaker 03

  So if you were running Google, competing against yourself, what would you do?

• 33:10Speaker 01

  I think, I don't envy that job at all. Nobody in the world wants that job. It's a very difficult job.

• 33:20Speaker 04

  Would you sacrifice the business model in order to get like, a new, get the next product?

• 33:26Speaker 03

  Or would you ship it as a separate product? Like, if you're Google, would you just build a separate thing that is the perplexity competitor and sacrifice the distribution advantage that you have in the short run?

• 33:39Speaker 01

  Yeah, I don't like generally, I don't know. I think I can say all of what I want, but they have more data on like what their users are doing. And there are a lot of people in the world who hate AI, by the way. So I think just throwing AI down people's throats on such a, you know, massive distribution area is not easy. What I would do, I would, I definitely don't know, and I don't want to be in that position also. By the way, if, if, if ads are part of every AI answer, you're gonna hate it too. And, and so it's good that there, there's alternatives like us.

• 34:14Speaker 05

  All right.

• 34:17Speaker 06

  Hey, everyone. My name is Akshat. So in a recent interview with Nikhil Kamath, he asked you for an internship at Perplexity. So I was just wondering how that arrangement is going.

• 34:28Speaker 01

  He, he came, he came to the office, he spent a couple of days. I mean, he hasn't posted about it, so I'll let him post about it. But I said, we did spend time with him. It was not a proper internship, but we did speak to him for a while.

• 34:42Speaker 05

  I want to start by saying thank you very much for your very candid answers. I really appreciated that. So a lot of startups, they find some like cool application of foundation models. And then they'll like build something off of that. But then if it does gain traction, then the foundation models will consolidate that into their own infrastructure. And Perplexity sort of has that issue too with like a lot of L L M's adding search like Chat G P T, Gemini, companies like Cohere. So I was just wondering like, how would you approach something like that? Would you try to pivot, just get better at what you do or?

• 35:18Speaker 01

  I think I would say pick something you wanna like be known for. Yes, there are other people integrating search, but we still want to be the fastest and most accurate. And obviously I cannot just say that and, and then stop. Like we need to figure out a new strategy too. And, and build new products that don't exist yet. So our browser will be that bad for us. And browser and search are not two distinctive products. They're actually, like the browser is a natural graduation step from search. Just like how Google graduated from Google Search to Chrome. And Chrome is the main reason they got billions of daily queries from hundreds of millions. So when Google IPO'd, they had no browser and they had like maybe a hundred million queries. Now you know like it's like ten billion or something.

• 36:04Speaker 01

  So the browser is an important part of that. And then so that's why we are making a massive bet on that. And agents can only be built with a browser. I am very like convinced about that vision that if you want to have a mobile agent that you can actually build and implement without being restricted by whatever OS rules that Apple or Google sets in terms of not being able to call third-party apps. Expecting every mobile app to have MCP servers and then like connecting all their data to your thing is not going to be that straightforward. Like, nobody wants to be disintermediated by an AI that quickly. So the browser will be a great way to build all these things. Thank you.

• 36:43Speaker 07

  So as a lot of us here have done, we've tried, we've failed at our startups. You know, some of us have been more successful than others, some like me have failed. When you're in that moment failing over and over again, what do you tell yourself as CEO or as an entrepreneur to win, to teach yourself to win?

• 37:04Speaker 01

  What do I tell myself when I feel like I might fail?

• 37:06Speaker 07

  Yeah, or when you're in that very specific moment of failing where you feel like everything's crashing down on you, or this feature isn't working, or this bug has popped up. How do you get through that? And what do you think your biggest motivational factor is in that realm?

• 37:19Speaker 02

  Or maybe like at the beginning, before it started to take off, what gave you the hope to keep working on it versus just go back to OpenAI and get your job?

• 37:29Speaker 01

  I just watched the Elon Musk videos on YouTube. No, I'm serious.

• 37:43Speaker 01

  I can tell you which video. There's a video where there's like a third failure in a row and like, what did you think? And he's like, I don't ever give up. I would have to be dead or incapacitated.

• 37:53Speaker 07

  So you'd say you're also never going to give up?

• 37:55Speaker 01

  Yeah, I hope to. I hope to like stay that way. It's not easy. I think he's done it for way longer. That's why you all like respect him. But that's, you know, there are examples of great entrepreneurs have done this despite all the odds stacked against them. So what do you have to lose to just keep going?

• 38:14Speaker 07

  Thank you.

• 38:24Speaker 08

  Yeah, my question is about kind of the sustainability of perplexity, not in terms of the business model, but just in terms of the web in general. You know, a lot of studies have come out recently showing that AI search engines like perplexity drive a lot less traffic to websites. So I'm curious, what do you think like the web will look like in five to 10 years? When a lot of these websites, you know, they're not getting as much traffic. And so they have to kind of cease their operations and like the web will just be a lot quieter of a place for content creation. How do you think perplexity fits into that? And what do you think the web will look like in that era?

• 38:53Speaker 01

  I think that there are gonna be, you know, the web is already pretty long tail. And there's a massive power law, so I feel like the power law is gonna get even more skewed. That is very obvious. There are gonna be certain brands that are well known and they're gonna preserve direct organic visits. But those who are trying to game the SEO system and trying to get traffic, I think they're definitely gonna have a harder time.

• 39:17Speaker 08

  Okay.

• 39:19Speaker 09

  Hi, Evan. Good afternoon. Firstly, where do you place the line between summarization and plagiarism in report generation? And how do you avoid IP violations in your product? And secondly, how do you deal with political bias and political, sorry, political bias and personal interest in news articles and other human written sources?

• 39:40Speaker 01

  Yeah, I think there are cases where you actually have objective truth, right? Like, what was the score in the NBA game? What is the live weather right now in San Francisco? Where you don't wanna, you don't wanna be wrong ever on those queries. And people know what is true. But you, even there, you're trusting, right? Like, you're trusting some data provider who's tracking the live game, the TV that's showing you the number, or Apple or Google's like AccuWeather, all these things. So at some point, it all relies on trust. And trust is built over time based on being being accurate reliably. And so trying to surface the right data from the right people who earned the right to like, like be surfaced in AI is how we think about it for accuracy.

• 40:25Speaker 01

  Now, there are things that don't have one clear, accurate answer. I think there...

• 40:32Speaker 01

  The best thing we can do is offer all the perspectives. And not really take a clear opinion on like what is right and wrong. When there is no clear answer to that question.

• 40:43Speaker 02

  Do you measure how accurate you are at that job by user feedback in some way?

• 40:49Speaker 01

  We don't actually measure it today. We should. And eval set should be built for that, like like questions where there is no one objective answer. The problem with building an automated eval for for that type of thing. Is what what is the right answer? It's subjective, right? Like like if if there are questions about the origins of COVID and there's so many different opinions of that. Relying a lot on Wikipedia as a source and and you know, can say maybe for a human raider. You're like, OK, I was saying all the things Wikipedia said it's a good answer. But maybe what you want is to say stuff that is not there in Wikipedia. And that relies on like having a much better human evaluator, like pool. Much smarter people who who are supposed to rate these things.

• 41:34Speaker 01

  And they're not like available for like, you know, scale AI style evaluations, right?

• 41:38Speaker 02

  Right. OK, I think we have time for one final question. You get it.

• 41:44Speaker 10

  Awesome. Hi, my name is Angela. Thank you so much for talking to us. I have a question about your go to market strategy. You had a great campaign for students. It's how I and assume many college students learn about you guys. But then also you had a collaboration with Calci, which is a little bit different audience. So I'm just trying to understand how do you decide which audience you're trying to get?

• 42:04Speaker 01

  I think like one one perspective here is trying to get into distributions of users that you don't typically have access to on your traditional marketing channels. You know, there are a lot of people who don't use Twitter or LinkedIn. And and and they're all like, they all exist and well, we just are living in a bubble here. And and there are some other businesses that have good access to them. Like, you know, traditional businesses. Like, you could imagine the kind of people who use Costco regularly may not even be using AI on a regular basis. And so if that's the kind of people you're going for, then it makes sense to change your strategy to reach them. But also remember that it's good to grow with adjacencies.

• 42:47Speaker 01

  Like, you do want to have some overlapping sets of people who would be the word-of-mouth carriers as they help you expand to more, you know, non-overlapping circles. So I think that's how I think about it. There should be some overlap, but your distribution should keep evolving over time. All right, Aravind, thanks for joining us. Thank you.