Perplexity AI CEO Aravind Srinivas, PhD 21, on why he ditched pitch decks

After years interning and working as a researcher at companies like Google, DeepMind, and OpenAI, Aravind Srinivas, PhD 21 (computer science), realized he needed to be his own boss. 

“Some people are unemployable. They just don’t listen to what the boss tells them to do,” Srinivas co-founder, president, and CEO of Perplexity AI, joked at a recent Dean’s Speaker Series talk. “I’m one of them.”

Watch the video of Srinivas’ talk:

Now, he said, Perplexity’s customers are the boss. “‘This is not good.’ ‘That’s not good.’’’ Every day I wake up, my phone has hundreds of these (messages) because I see across all these different social platforms, and that’s how I know, every day, there’s some work to do.”

Founded in 2022, Perplexity’s chatbot search engine now logs more than 300 million queries a week. With a mission to “serve the world’s curiosity,” the company recently released Comet, an AI-powered web browser that integrates an AI assistant directly into its users browsing experience. 

Srinivas’ path toward entrepreneurship began during his PhD in computer science at UC Berkeley. He was drawn to the idea that both business and computer science start with an idea that you have to convince others to believe in, he said, whether that’s through writing a research grant or pitching to investors.  “I’ve always looked up to entrepreneurs,” he said. “Driving change the way you think should be done can only be truly done as an entrepreneur.” 

Srinivas said what distinguishes Perplexity from legacy search engines isn’t its infrastructure but its approach toward search. Rather than rely on keywords, ads, or clickbait, the platform aims to answer questions with direct, reliable answers. This question-answer philosophy carries into the Perplexity workplace, too, where Srinivas said he starts meetings with questions, not presentations. He uses the same approach with investors.

“I’ve never done a pitch deck for any of the other Perplexity funding rounds,” he said. “I just write a memo, and I tell them, ‘You can do a Q&A and ask whatever you want. I’ll spend two hours with you, ask me all the questions you have, and we’re just going to pull the metrics for whatever you have and show you right there. And anything else that is internal, not internal data, you can ask Perplexity. It already knows everything.’…I’m not exaggerating.”

Throughout his time as a tech leader, Srinivas said he’s learned to accept imperfection. “I used to think every problem just needed to be fixed instantly,” he said. “As the company has scaled, I have learned to make peace with some problems just existing. That is the number one skill you need as a CEO or a founder.” 

By releasing products that are 80% perfect, Perplexity has been able to adapt to the quickly developing AI market, he said. Each release becomes a foundation for improvement as his team continues to refine products over time. 

To Srinivas, resilience also means redefining failure and never giving up. He said too many people avoid risks because they worry about losing or how others might perceive them, when in reality, most potential critics aren’t paying attention. “There’s really nothing to lose from taking risks,” he said.

Listen to Srinivas’ talk on the Dean’s Speaker Series Podcast:

TRANSCRIPT

ALEX MAS: What a turnout. I’m Alex Mas. I’m here—I’m associate dean for academic affairs here at UC Berkeley Haas, and I’m here stepping in for Dean Chatman, who was called away. I am so thrilled to have Aravind Srinivas here today. Let’s give him a huge welcome. And also, welcome back to Berkeley. Aravind obtained his PhD in computer science here at Cal, home of many luminaries who developed the foundations of machine learning in AI, such as Professor Pieter Abbeel and Michael Jordan. A couple of facts about Berkeley. I’m speaking to an audience of one right now, but UC Berkeley is the No. 1 university in the world in venture-backed startups, according to PitchBook. Stanford will say they’re the No. 1 per capita, to which I say, who cares about per capita. We do impact at scale here, and scale is what Perplexity is all about. They’re doing AI at scale. As CEO, Aravind leads Perplexity with a user base of over 30 million active users and 780 million inquiries and counting. Amazing for three years. I’ll just say I am one of those 30 million users. I’ve used the Perplexity Voice Assistant, and it’s just such a great product. Aravind is an exemplar of Questioning the Status Quo, as well as other Defining Leadership Principles. In a recent conversation with Yamini Rangan, CEO of HubSpot and a proud Haas alumna and board member, Aravind shared that curiosity is the key to success in business. It’s what drives innovation and progress. I couldn’t agree more. It exemplifies what we call at Haas the human edge of innovation. We’re training the first generation of AI native business leaders who understand that AI’s true power lies not just in technology, but embodying the deeply human skills of curiosity, judgment, and creativity. It’s about asking better questions and using AI as a tool for meaningful innovation, keeping humanity at the center of technological progress. We are so thrilled to have Aravind with us today, and I’m going to turn it over to our excellent panel who I’d like you to just do one round quickly to introduce yourselves and what you do here at Haas. Thank you so much. 

GONZA VÁSQUEZ: Hi, everyone. Hi, Aravind. I’m Gonza. I’m a second-year full-time MBA student from Peru. I also happen to be one of the Haas Tech Club co-presidents, and it’s an honor to be here with you today. 

ARAVIND SRINIVAS: Thank you. 

REY YRUEGAS ALMANZA: Hi, everyone. My name is Rey. I’m a full-time MBA student from Mexico, and you have no idea how excited we all are. We’re all avid Perplexity users and Comet users as well. So, we’re just really grateful for you being here. 

ARAVIND SRINIVAS: Thank you. Firstly, I’m super excited to be here. I wish I did this before I did the Stanford GSB event, but better late than never. Back to the OGs. Yeah. 

GONZA: Aravind worked as a researcher for OpenAI and DeepMind, and for a lot of people, that would be a dream job. But you choose to leave that space and start your own thing. So what was the signal or realization that made you take that leap? 

ARAVIND SRINIVAS: Well, some people are unemployable. They just don’t listen to what the boss tells them to do. I’m one of them. So that’s the true reason why. What would it look like if I didn’t have a boss so that I could—of course, for your own accountability and staying sane, you need to have something to keep you under check. And for me, it’s the users. Whatever they tell me, ‘The product sucks.’ ‘This is not good.’ ‘That’s not good.’ Every day I wake up, my phone has hundreds of these because I see across all these different social platforms, and that’s how I know, every day, there’s some work to do. You’ve not made it yet. And I keep going. But yeah, jokes aside, I’ve always looked up to entrepreneurs. I think the spirit of driving change the way you see should be done can only be truly done as an entrepreneur. One of my entrepreneurial inspirations has been Larry Page, the founder of Google. And there is a particular thing he said, which is, ‘If you truly want to do something very impactful in the world, there’s probably only two career paths that you can. One is an academic. You become a professor, you have a lab and you get some funding. You get great students, and you try to change the world through your research, or you start a company.’ And pretty much the same thing. ‘You go to VCs, you ask for funding, you try to get a good group of initial founding team and drive the change that you want to bring in. Any other thing, it’s pretty hard to do in a way that you want to do.’ And I definitely didn’t want to be an academic forever for many reasons. Obviously, one is it’s pretty hard to actually get funded, especially in the current climate. So I think startups was something I definitely look for doing. 

GONZA: And what did you struggle the most with that transition from researcher to CEO? 

ARAVIND SRINIVAS: Interestingly, it’s not as disconnected as it looks like. And that’s the thing. When I explain to people, they get it. But on the surface level, it doesn’t make sense. Both research and in startups, you’re essentially trying to convince people to work with you. You have to start with a contrarian insight, an unconventional idea. No one wants to read a paper that’s telling them obvious things. And reviewers wouldn’t accept it in an event—in a conference, anyway. You have to write grants for your professor at some point. First year, university covers you, but second year, third year, you do have to help the professor write grants and bring in funding. Same thing. You have to talk to VCs and help you get funding. If you want to do bigger projects, you need to have junior students also help you there. You need to hire people and founding team to help you up. You have to market your research. It’s not important to just get a paper accepted in a conference and go do a poster or an oral presentation. You actually need to have other people building on your work. That’s how you gain your currency in the academic world. Citations. If your paper is not cited by other people, especially important people, your papers are not that useful. Same thing in startups. Your products, whether it’s consumer or enterprise, you need to have customers. You need to have users. And then you need to build on top of what you have. You can’t stop with one paper and then move on. You need to have a line of work. That’s how you make your name in the field. Same thing with startups. In fact, I would say in startups, the line of work, the incrementality, the additional work that keeps going, the iterations, is even more. But the last thing, the most important thing, there’s a lot of suffering in both. Delayed reward, the concept of delayed reward. You are writing—you’re working on a line of research. You have no idea if it’s going to work. If you already knew it, there’s nothing interesting in it. Same thing with startups. You have a thesis of why this could be an interesting product to build. No one knows if it’s really going to work. And so that concept of dealing with suffering and delayed reward and having that intrinsic motivation, everything is built in a research career. So I would say I was actually very well-prepared for startups by being in academia as a researcher. 

ALEX MAS: Aravind, we have a little bit of a complaint for you. Basically, when we were preparing this chat, Gonza and I sat down and did a bunch of research. And we said, ‘We get it now. We know what to talk about.’ However, Perplexity would launch a new product. And then Perplexity would buy a new company. So I think our question is, how do you lead such an agile and ever-changing organization? 

ARAVIND SRINIVAS: Well, firstly, necessity is the mother of invention. I don’t have a choice. It’s pretty brutal out there. So just to survive in this fast-moving AI world, we have to keep moving fast. We have to ship a lot of things. We have to constantly—so the most interesting thing in AI, which is very different from other eras, is previous times, you would have a pretty stable infrastructure and tooling to build around. And so you can build around that. You don’t have to think about, ‘What would be the infrastructure six months from now?’ and try to build for it today. That’s the difference in AI. You have to build imperfect products, not perfect products. You have to build products that 80% work. It cannot be 60% because no one’s going to use it. The retention will be zero after 12 weeks. You have to build 80% perfect products where there is a long tail of 20% that kind of doesn’t work, but that 80% is enough for users to get excited about it. And six months from now, that 80/20 goes to 90/10, and 12 months from now, it goes to 95/5. And that’s how you win the market one or two years from now. You have to start with something imperfect today, and you have to have a vision for where it will be, like the state of AI will be six months or 12 months from now. So that’s always been the hard part, and that’s what we

tend to focus on. For example, everyone’s talking about our Comet browser, which we made generally available last week. It’s a world’s first agentic browser. You can go and do stuff for you, pull in your personal context and stuff like that. We started to work on it one year ago when we decided to buy a small browser team and start working on agents. Agents was such a buzzword

for two years prior to that. No one cracked it. Everyone thought, ‘It’s just a buzzword, nothing’s going to work.’ But we knew that reasoning models were beginning to make progress. OpenAI released the o1 model. O1 was laughed at. Everyone made fun of OpenAI for that. But we knew that three or four iterations from now, it will get there, which is where it is today. But you cannot start working on the browser today. You’ll be late. You would launch it one year later, and by the time, it’s too late. So we started working on it last year. So now we got to work on things that one year from now will be great, and I don’t quite know what it is yet. So yeah. 

GONZA: You mentioned in a previous interview that the way you inform your vision is you say, ‘I don’t want to know what’s going to change in the world in the next five to 10 years. I want to know what is not going to change in the world in that period.’

ARAVIND SRINIVAS: ‘To be rigorous.’ That is from Bezos. He said it. I liked it, and I’m saying it, too. Yeah. 

GONZA: Thanks for highlighting the copyright. What is it that you think is going to be that it’s not going to be a change for? 

ARAVIND SRINIVAS: Yeah, so that’s kind of actually what shaped our product values and also the company culture is speed. No one’s going to ask for a slower product, even if the answer is complicated, even if you’re prompt, is so complicated that it would take two hours or a week of research for a human to do, people still want quick, fast answers. And faster the product, the lower the infrastructure costs. And so it’s going to increase your margin profile, too. So there’s no questions asked. No one’s going to ask for a less accurate product. Hallucinations are always going to be a problem. Hallucinations should always be a bug for a product that aims to give you accurate answers or do stuff for you. You don’t want an agent that makes mistakes 20% of the time. And no one’s going to ask for a crappy user experience. Everybody wants a delightful UI. It should feel magical when you use the product. So these are things we focus on, like speed, accuracy, and world class UI. I know everybody wants the product to work for them, their personal context. No one cares if some tool has an MCP server or not. No one cares if you have a partnership with that company or not. That’s all for you to figure out. The user doesn’t care. They want it to all work magically across. And they want the product to work for them. Everybody—people want the product to help them feel even more at a convenience, comfortable, instead of writing really long prompts. People are not going to come to you and say, ‘I want to write a 20-line prompt, make your product work for that.’ In fact, they’re going to tell you, ‘Hey, I want to shorten my prompt, and I want you to do the hard work of really expanding it, understanding it, asking me clarifying questions.’ So these are the areas that we’ll be working on for the next five years that nothing changes there. 

GONZA: Perfect. 

ALEX MAS: Aravind, Perplexity’s mission is to support the world’s curiosity. Let’s pretend we’re in your performance review. Aravind, an enhancing curiosity is your main OKR. How do you feel about it? How are you doing? 

ARAVIND SRINIVAS: I haven’t done a performance review in a while. I hope not to have to do it. But yeah, KPI in terms of what measures, how well we’re doing on our mission is how many questions a day are we able to answer. How many questions are being asked on all product surfaces of Perplexity? And that’s growing really fast. I publicly mentioned we are doing in excess of 300 million queries a week. So I think the professor mentioned 780 million queries a month. That’s already an old number. And we’re doing well in excess of 300, actually. And that’s our primary metric. And our goal is to get this to billions a day. I think people do 10 billion searches a day on Google, something like that, across the world. I believe that it’s way more natural for people to ask a question. You don’t—we’re doing a Q&A session right now. You’re not asking me, ‘Aravind, KPI, Perplexity.’ You’re not asking me these keywords. That’s how you use Google today. Instead, you’re asking, ‘What is your KPI for Perplexity to measure how well you’re doing on your mission?’ That’s a question. That’s way more natural. When you can do that, when there is a tool that can give you an answer for that, instantly, you’re going to do more of it. Because every time you get an accurate answer, it leads you to the next question, which is why in our UI we actually suggest you to ask the next set of questions, because our goal is not to stop with that answer. It’s to let you stay curious and keep asking more. And so, we think the TAM is actually even bigger than the Google search query volume today. 

GONZA: Do you believe that just putting out these super smart AI tools will help people be more curious or smarter? We were discussing about that with Rey. How much is it the technology itself and how much is it the change that you want to show? 

ARAVIND SRINIVAS: Yeah, I think it needs to be both. The technology should first work. If the product is not accurate, if the product is not fast, it’s not fun to use, it’s not trustworthy. So you’re not going to trust it. So you’re not going to ask questions. And so, that’s automatically going to not make you more curious. But even if the product is great and it works, it’s fast, accurate, fundamentally, the way we do work needs to change. Earlier, you would be considered smart if you came into a meeting, and you appeared like a know-it-all, right? You knew everything. Now, I think what’s going to change is you’re going to ask the right set of questions. It doesn’t matter what you know. There’s an AI that always knows more than you on any topic. It doesn’t matter how smart you are, the AI is going to be also able to figure it out. But where you’re going to have your edges, even being curious about the right things, posing the right questions, rethinking an existing situation through a different lens, connecting dots between things that don’t exist. The relationship between two things that don’t exist. For example, you asked me this question of researcher and startups. Not many people understand that connection well. So that’s the kind of way in which we can have a different kind of work, we kind of call it the question work. Meetings can start with questions and meetings can end with a new question. That’s how we skip a lot of meetings in between, which have a lot of presentations and slide decks and templates and people sharing templates, et cetera. Famously, I never did—after the Series A was the only time I made a pitch deck. I’ve never done a pitch deck for any of the other Perplexity funding rounds. I just write a memo, and I tell them, ‘You can do a Q&A and ask whatever you want. I’ll spend two hours with you, ask me all the questions you have, and we’re just going to pull the metrics for whatever you have and show you right there. And anything else that is internal, not internal data, you can ask Perplexity. Like, it already knows everything.’ This is—by the way, I’m not even exaggerating. The most recent funding round, I think at the time, one of the investors who put in a lot of money, I did a Zoom call, a webinar, and I answered a lot of questions. And then there was a lot of follow-up questions from there that they sent in a long email. And I copied the entire email, put it into Perplexity, and said, ‘Answer it like Aravind.’ And it has all the context of whatever I’ve said in all the YouTube interviews and different blog posts. And I took it out, and I saw it. And it was like, ‘Damn, I don’t think I could have done as good a job as this.’ So I actually just replied to that email with the Perplexity answer link, and I asked, ‘See if this suffices. If not, I can add more context.’ And then they said, ‘This is wonderful, and they wired the money the next day.’ So it’s a free product that helps you raise quite a lot of money. 

GONZA: And have you been doing that for CEO tasks? 

ARAVIND SRINIVAS: Yeah. I mean, a lot of the CEO tasks is essentially information gathering. And then analysis of whatever information you gathered. And then posing the right set of questions on top of that, and then making decisions and repeat. Except the first four steps are now very easy. If you have a browser, a Comet, it has access to your Slack, your Google Workspace, all that stuff that you have on your browser, and the web. So the information gathering part doesn’t need to be done by other people. I can just do it myself. And then making decisions is still my—that’s why I still have a job. Though I do take the help of Perplexity when I need to make hard decisions. And then just communicating that, which I still do myself, and following up in case things are going slower than expected. That’s how I do the job. In fact, I really think I would love for all the end-to-end thing to be done by an AI so that I can literally just take vacations because it’s been a long time since I took a proper one, but I feel like that’s still quite some time away. Yeah. 

REY: One of the things that we were curious about was, going back to the question about moving so fast, and how you mentioned really the industry demands that of you. How do you make your team keep up with your pace personally? 

ARAVIND SRINIVAS: One of the things we say internally is, evaluation is growing really fast. Beginning of last year, we valued at $520 million. Today, we valued at $20 billion. So that’s like a 40x jump in less than two years. It’s impossible for a human being to grow at 40x in less than two years. Yes, we believe in the mantra of 1.01 to the power of 365 is 37.78. So if you improve 1% a day you get 3,700% better at the end of the year. But humanly it’s hard to compound like a machine. People need some time. They get burnt out. So what I believe is, try to take on increasingly complex projects. As the company grows, even doing the same task becomes more complex because you have more users. So earlier, we would just YOLO ship things. I would just ask for a feature, and they would code it up and just ship it to production. No questions asked. Who even cares. You have 1,000 users, 10,000 users. Now, there are so many DAOs, like tens of millions of DAOs, and you have to think. And a moment there’s even a small mistake, people will screenshot it and shame us on Reddit. And it’s very, very hard to read some of those things. And there is a churn in revenue. People start canceling the subscription plans if you make mistakes. So at the same time, our edge comes from kind of ignoring all these side effects of shipping fast and still moving at the pace we do because that’s what we can do that Google cannot do. Because for Google it’s $1 billion, 100x harder problem to solve. So I still try to tell them, ‘Be a little more responsible than you were two years ago, but still move fast and take on more complex projects. If you feel bored about what you’re working on, try to find slightly different things to work on that give you some dopamine. Burnout usually comes when you’re—it’s less about the hours. Burnout usually comes when you’re not enjoying the work you do, but you’re actually having to do a lot of it. And so, that’s essentially you having to listen to your own internal dopamine system. And if you’re winning, the metrics are all going well, that essentially drives you to put in the hours. 

GONZA: So in the last couple of years, you have been shipping a lot of products. You’ve added Pages, API integrations. More recently, have the email assistant and of course, Comet. Each of these products represents a different problem space and challenge. So if you could add a magic wand that solves just one problem for Perplexity’s user today, what would that be? 

ARAVIND SRINIVAS: Context. The user context system is the most important thing. That’s why everyone’s working on browser, memory, and all these things that truly understanding the user so that every answer is personalized, actions are taken on your behalf, things can run in the background. You’re like all the boring aspects of your work and life can run on Comet. That would be fantastic. And a lot of these problems we don’t know how to solve. I want to be able to be on a walk talking to you guys about random things about campus and still be summoning some tasks on Comet. It’ll be working for me. That way my life feels so easy and gives this power to everybody else. And that’s how you change the world. And doesn’t seem easy to solve a lot of hard engineering problems, because doing the browsing just on the client side is so much work. Doing it on the server side is even more work. Doing it in a way that’s synchronized across all your devices different platforms is even harder. And doing it in a way that’s proactive so that you don’t even have to ask the AI to do stuff, it automatically knows what you need and does it for you is incredibly hard. And we want to solve all these problems. That’s how you deliver the ultimate magic. 

REY: Aravind, what do you think is an aspect of AI that is underestimated? A lot of people talk about energy, GPUs, cost of inference. But what do you think is something that is not being talked about enough?

ARAVIND SRINIVAS: I guess the same thing. Orchestrating all these different tools in a seamless way is considered, or rather, people just don’t even think about it. People talk about, ‘Oh, you’re building a model, you’re building a rapper.’ That’s the level of distinction they look at. But AI is no longer just a model. It’s a whole end-to-end system. Just the model itself doesn’t know how to use the web, doesn’t how to browse, doesn’t how to use tools, doesn’t know how to use imperfect tools. We humans are great. For example, LinkedIn is a tool I’m sure everyone here uses. Search engine on LinkedIn is so bad, and we humans still figure our way around it. If you’re trying to source people for your next startup, you’re trying to find, and you have a rough sense of these are the companies I want to find from. So then, you look for people within your network there. You have all these selectors, these filters, dropdowns, and you somehow figure your way around. AI doesn’t that—LinkedIn search sucks. So it has to learn its imperfections and work around it. And LinkedIn is just one tool. We use hundreds of tools on a daily basis to get our work done. So if you truly want to build an AI knowledge worker, it has to work with the imperfections of the human world and still go do stuff for us. So that is an end-to-end system that pulls context across tools, works with imperfections, and reliably does the work for you on the background. People don’t realize how hard it is, and I think the system is way more challenging than just training another model. That’s why as a company, we decided, ‘OK, it’s too late to work on a model.’ It’s going to cost you $100 billion, whatever. It’s not just about one release. You have to keep doing it. But in order to build a really world-class personal assistant or an executive assistant or a recruiter or a saucer, you need the system. It’s not just the model, and it needs a browsing capability. It needs a search, indexing, crawling, tool use, orchestration, memory. So we go work on all these other valuable problems. And that’s not talked about as much, in my opinion. 

REY: On the value propositions that you mentioned, you have been heavily investing in going international with India’s Financial Tool. What’s the strategy behind that? 

ARAVIND SRINIVAS: Well, this is a tool that’s being used by people across the world. I think all AI companies, at best, the American market share is at best 25%. Not more than that. So 75% of the users is coming from outside America, split across Europe, Asia, and then South America and so on. So definitely, you need to focus on that and language support and local use cases, verticals. Some verticals matter in different industries. You have to do all the work. That’s how you go from an interesting tool to a ubiquitous tool. So hence why we are focused on that. 

REY: So, worth thinking about. These are some of the world’s hardest problems to solve. And you mentioned in our interview that one of your mottos or how you stay strong is if you say, it’s over only when I say it’s over. So how did you build that confidence over the years? 

ARAVIND SRINIVAS: It’s not a confidence. It’s more grit. Why do you need to give up? Keep trying. In fact, this is just an anecdote. Rafael Nadal, I’m sure you all know what he’s known for, the determination. He plays like every point is the last point. So he was practicing in Indian Wells, the tennis tournament. And Larry Ellison owns the tournament, I think. So he was hosting Nadal, and they were just practicing on the court. And Nadal asked Ellison, ‘Hey, what is one advice you would give me?’ It’s pretty rare for Nadal to ask someone for advice. And Ellison goes, ‘Oh, you have to be contrarian. You have to think of ideas that other entrepreneurs are not thinking of. You have to go do it fast.’ And he goes on a whole long list of things. And then he’s like, ‘Hey, Rafa, hold on. Ignore everything I’ve said. You just need to never give up.’ And then Nadal was like, ‘Yeah, tell me about it.’ But that’s the bottom line. It doesn’t matter. Entrepreneurship, sports, anything. I think the greatest have this one quality that, across all of them, is they don’t ever give up. 

ALEX MAS: Reflecting on your journey, what’s one belief that you held early in your career that you just no longer hold? 

ARAVIND SRINIVAS: I haven’t actually thought about it. But let me try to think and give you an answer. I guess I used to—I mean, I still have that thing. So I see it being helpful to me, but I guess I used to think every problem just needs to be fixed instantly. And as the company has scaled, I have learned to make peace with some problems just existing because I know it’s impossible to fix every problem. And a corollary of this is that anytime you ask someone, ‘What are you doing, what is the one problem you’re looking at,’ or, ‘What are your priorities for this week?’, they will have a bucket list of like five to 10 things to say. And then, I asked them, ‘OK, what is the number one thing?’ And they can’t say it. So the reason they can’t say it is they don’t force themselves to pick one. And when you’re forced to pick one, you just become indecisive. It’s hard because you’re like, ‘Oh, what about the rest?’ That’s also important. And I think that is the number one skill you need as a CEO or a founder is anytime you need to pick the number one thing. Obviously, just doing one thing is insufficient in this kind of fast-moving world, but you need to know what is the number one thing and just learn to let other problems stay there for a while. I didn’t operate like that at the beginning. I used to want to fix everything right away. Now, I’ve learned to operate slightly differently. 

REY: So going back to the browser space, you have been compared to that other big tech company that focused on search, that starts with AG. 

ARAVIND SRINIVAS: I haven’t heard about it. I heard they were founded at Stanford. Do you that? Yeah. 

REY: What do you think is broken or to be improved on the traditional search experience? And why did you say, ‘OK, we’re going to build an AI browser?’ 

ARAVIND SRINIVAS: Well, let me zoom out and think about it as, what is our mission? Our mission is to serve the world’s curiosity. And the best way to do that is

to let people ask questions. Is the current internet designed for that? Not really. There’s a lot of fake news, clickbait, ad content. Anywhere you are on the current web, you’re on someone else’s funnel. So someone has to step back and change it. And you cannot really build a truly personal assistant that actually helps you to delegate a lot of the boring things so that you can go explore and investigate and dig deeper in interesting things without building a browser. You need to reimagine the whole thing from the get-go. And that’s why we wanted to build it. We didn’t want to build a browser just to put our search engine there. We think our search is superior to Google’s, mainly because we think the world is moving toward questions from keywords—not because we have much superior technology and infrastructure, I think

they’re still ahead of us there. But the thesis is that people are going to want questions. People don’t want clickbait. People don’t want ads. People don’t want SEO slop. So that is fundamentally changed in Comet. And people want to own AI themselves. Right now, when you’re in Google’s world, they own all the AI. You don’t have anything. So they tell you what is a ranking. They dictate who gets to rank here and there. They get to dictate what ads you see, and you just take it for granted. With Comet, it’s very different. You have a personal assistant. You can actually say, ‘Hey, go Google these 20 queries for me. Ignore all the ads. Click on the most relevant things for me based on what you know about me, and populate this in a Google sheet and send it as an email to my colleagues.’ Google will not allow you to do this because they lose money if you do this. So that’s why we come in. 

ALEX MAS: Now Aravind, this will be our last question, and then we’ll open it up to public Q&A. And the way we want it to end was basically we wanted to create a hallmark question for the speaker series, and we thought we would start with you because you’re so close to us here in the community. And the question is targeted toward getting to you better as a person rather than just the professional and your achievements. The question is, if you had one extra week in your life, what would you use it for? 

ARAVIND SRINIVAS: Honest answer is I would still work. Yes, let me just be honest. But yeah, I don’t get much time to spend with my parents, so I would obviously try to spend with parents. I would still work then, but I would go and spend that time with my parents.

ALEX MAS: Thank you. Aravind, Rey, Gonzalo, thank you for such a great conversation. We have a lot of questions, and I’m sorry to the audience. I don’t think I’m going to be able to get them all, but I’ll do my best. So I ordered them more or less randomly, so we can just fire off. Question one: What do you consider as the moat for Perplexity? 

ARAVIND SRINIVAS: Speed. 

ALEX MAS: Great. I like that. This is a longer question. As AI-driven workflows replace traditional processes, how should leadership evolve, particularly in big companies? A lot of our students here are going to be leaders in companies, and they may not be doing AI, but they’ll be using AI. So in decision-making or design talent development. And where do you see the biggest leadership impact over the next 12 to 24 months? 

ARAVIND SRINIVAS: Evolve to doing question work. Start meetings with questions, not with presentations. Skip the first meeting. Get to the second meeting with starting the question. And incentivize people to be asking questions rather than trying to know answers to things.

ALEX MAS: You’ve named several personality attributes critical to being a founder, including grit, which really resonated with me. So when thinking about your team, what are complementary personality traits, different from your own, that are essential for team’s collective success?

ARAVIND SRINIVAS: Well, all of them are way smarter than I am. They—my co-founders—are way better engineers than I am, more tactical at getting things done, better project managers, better deal makers. So yeah, this is just meant as a joke, but you have to be a jack-of-all-trades as a CEO. And you can go deep into things if you want to, but you have to

be a Jack of all trades. So sometimes you might internally feel you’re just a bullshitter, but the others are just way better at actually doing things than you are. 

ALEX MAS: Just a follow-up question. Was there a transition point for you where you had to—what was the key point where you moved from pure startup mode, where you felt like you were doing everything, to where you had to take that one step back and form the team?

ARAVIND SRINIVAS: I realized this from the get-go that without a good team, we’re not going to go that far. And we—one thing I really appreciate working in Perplexity, leave alone as a founder or CEO, is how talented everybody is at doing their work. They’re world class at whatever they do. So just learning from them is fantastic. And I always like learning from others. I think there was no insecurity for me that someone’s better. I actually always wanted to be around those people. 

ALEX MAS: When do you expect agents to have autonomy to make transactions on users’ behalf?

ARAVIND SRINIVAS: It’s already happening on Comet. You can actually have Comet even trade for you right now. As long as you’re logged into your Robinhood and have it as a tab, you

can have Comet play call options on Nvidia and stuff. Well, I’m not recommending that by the way. 

ALEX MAS: Well, actually, it’s relevant. That question was from Ian Kaufman, who’s a MASter of financial engineering student. So that’s maybe what they had in mind. What about regular consumer transactions, and that’s—

ARAVIND SRINIVAS: Yeah, it could shop for you. It can book a United flight on your behalf. It can apply a coupon for you to check out so that you get a discount. All these things. And we’re going to make it more seamless and smooth. There’s some amount of friction right now, but as long as you’re having a wallet on the browser, it’s like Chrome, the agent can do all these things.

ALEX MAS: How do you look at failure when chances of failure are higher or even guaranteed? How do you still perform and show up? 

ARAVIND SRINIVAS: What is there to lose? A lot of people operate in this mindset that there’s so much to lose and, ‘Oh, what will people think about me if I lose?’ And like, no one even cares. There’s just so much of problems in everyone’s lives already. So the only person who’s going to support you anyway is your family. And as long as they support you, even in the time of failure, which that’s what a family is meant for, and close friends. There’s really nothing to lose from taking risks. 

ALEX MAS: How does Perplexity approach data privacy and IP issues in AI, and how has this changed over time? 

ARAVIND SRINIVAS: Yes, so we really give the user clear instructions. You only connect your different MCP connectors if you really want to. There’s a value add from that. And all the enterprise usage of Perplexity is zero data retention. So we don’t actually even look at your data. The admin account gets to store it, but that’s for internal admin usage. We don’t actually get to see it at all. And we support incognito mode searches both within the product, on the browser, too, and none of that data is ever stored for training purposes. All the data that lives on your browser on the client side just stays on the client. It’s not sent to the server in any way, unless you invoke the eye and ask it for help. The browser just stays as a regular private secure browser. 

ALEX MAS: I know we have entrepreneurs in this room. Relevant question for someone starting out. So if you’re raising from VC angels for the first time, what are some red flags to spot in investors? 

ARAVIND SRINIVAS: Wow. Where do I start? I mean, I guess I would just say, look for—I mean, this is usual advice. Look for a long-term investor. If they’re asking you, ‘What’s the unit economics?’, this that at the seed stage, clearly they’re not the right fit because no one knows the business plan. No one has any idea what the actual product that you’re going to build will be at the seed stage. So you’re really betting on the founder, the team, the mentality of the founder, the skill set, the ability to sell a story. That’s what you’re betting on. The actual product, the actual roadmap, everything changes pretty fast. So look for people who are OK with that and are betting on you rather than your idea or the unit economics or whatever. Because if the idea is right, they’re going to bet on someone else who’s doing the same idea too, because they’re more interested in that. 

ALEX MAS: How—there’s a lot talk about the environmental impact of AI in general. And I know there’s a lot that’s known and not known. But how do you think about it at Perplexity in terms of energy use, servers, and also, if your vision of growth kind of continues in the future as well? 

ARAVIND SRINIVAS: Yeah, so we’re not in the business of building data centers. In fact, a lot of our inference usage is pretty compact and efficient. We try to use open-source models whenever we can. We work a lot on distillation. We publish a lot of blog posts on how we can do inference even more efficiently of trillion-parameter models with minimal GPU nodes. And we really track the token throughput and efficiency. So I always believe incentives matter more than interests fundamentally for a capitalistic company. Our incentives are actually aligned with being efficient because search is a business where you have to be fast, and you have to scale to a billion queries or more. So without being efficient, we’re going to lose a lot of money. We’ll be out of business pretty fast. So hence, by working on efficient infrastructure, we will do our part to be responsible in terms of energy usage. 

ALEX MAS: Efficiency and sustainability often go hand-in-hand. 

ARAVIND SRINIVAS: Yeah. 

ALEX MAS: OK, should I ask this question? I guess I’ll ask this question. Are we in an AI bubble? What are your thoughts on that?

ARAVIND SRINIVAS: I mean, I think it’s interesting. There are obviously announcements you hear and people without any product or something can raise $1 billion. So that does feel pretty odd. I raised the first million dollars, and that itself was not that easy. This was three years ago. So definitely there is some bubble. Value is being created though, otherwise people are not going to be using the products this much. So it definitely feels very different from crypto, where there’s imaginary asset value creation in these assets, but there is no real tangible usage. So by building a personal assistant and putting it in the hands of everybody and making their life easier, making them work more efficiently, more fun, I think we’re creating value in the lives of people, and that will automatically translate to economic value. 

ALEX MAS: So as your company has been so successful and you as a founder and CEO have now become so well-known and famous as a result, how do you stay grounded? 

ARAVIND SRINIVAS: I told you. I wake up and there’s 100 bugs that—today, morning, the Reddit post that got sent to me was ‘Perplexity sucks.’ Or ‘the browser is so hogging my RAM. You guys are supposed to be better than Chrome.’ So I have enough of these every day to deal with, and I think there will be a lot of this coming for the next five years. I’m not worried about staying grounded. 

ALEX MAS: Do I have time for one more? I think we end at 1:20. I’m looking at Sarah. I don’t where Sarah is. OK, well, I think I have one more. For a B2C product, what’s the most effective way to market the product for traction? Well, surely it must depend, but what’s your experience with that? 

ARAVIND SRINIVAS: Product-led growth is what we have mostly done. So we ship a lot, we keep announcing stuff. And so, even if people don’t use a new feature we announce, we’re always in the news. We’re always doing stuff, and we’re constantly keeping people interested and excited and bringing in new users. And then product obviously has to be really amazing. So that is what leads to word-of-mouth growth. And that’s how you get new activations. We’ve done a lot of partnerships where other companies with access to a lot of consumers would market our product, and in return they get to do some revenue sharing and things like that. So we partnered with a lot of telcos and other established brands here like LinkedIn, PayPal. So you got to keep doing a lot of these things to get new users. It’s very hard when you don’t have distribution. That’s why everybody discourages you from working on consumer because the last real consumer company like, of course, OpenAI exists there, but you could argue that they started the brand 10 years ago, and they founded with Elon Musk as a founder. But the last real company that did not have any of that large funding and was an unknown name that got a lot of users is TikTok. And TikTok spent billions of dollars on Instagram ads to get users. So it’s going to be a long road. If you really want to go to consumer, you have to absolutely want it and give it your best in all these different channels. 

ALEX MAS: Well, I did not think we’d get through the questions, but actions speak louder than words. True to form, the branding of speed carried the day here. So we appreciate that and all the opportunities to answer our questions. Huge round of applause for Aravind. 

ARAVIND SRINIVAS: Thank you. Thank you.