
15 segments available
YC General Partner David Lieb sits down with Aravind Srinivas, the co-founder and CEO of Perplexity, to discuss his origins in Silicon Valley, what it's like to compete with Google, and what the future of search could look like. Apply to Y Combinator: https://ycombinator.com/apply Chapters (Powered by https://bit.ly/chapterme-yc) - 0:00 Intro 0:51 Aravind’s early days in AI 3:45 What made Aravind start a company 6:35 The first iterations 14:10 Realizing Perplexity had potential 19:02 “The user is never wrong.” 22:20 Managing the team 24:53 Brutal honesty brings out the worst bugs 25:33 Avoiding becoming a big slow company 27:37 The future of search 31:11 Perplexity’s advantage against its competitors 34:26 Outro
Aravind Srinivas shares his journey into the world of AI, detailing his early experiences as an intern at OpenAI and the pivotal moment when he learned about unsupervised learning. He reflects on how this knowledge shifted his focus from reinforcement learning to generative models, setting the stage for his future endeavors in AI and search technology.
"we release the ability to ask follow-up questions that double the engagement time on the site and also increase the number of questions every day so I was like okay there's something here it's not wor..."
Aravind discusses the motivation behind starting his company, emphasizing the ambition to create a product that could rival Google. He explains the importance of query reformulation and how generative AI could enhance search engines, revealing insights gained after launching their product that shaped their approach to competition in the search market.
"that'll make the product even better and that'll lead to more users and more usage will lead to more data points and it'll become a flywheel and uh it should also be on the AI completeness path it's s..."
In this segment, Aravind recounts the initial iterations of their product, including the challenges faced in securing data from potential partners. He describes the development of a Twitter-based search tool that allowed users to explore social interactions, showcasing the innovative spirit and rapid development that characterized the early days of Perplexity.
"made sense this is again after launching the product we realized this not before um so there's like some benefit to being ignorant ignorance is bliss uh is that if people stop clicking on links the ad..."
Aravind explains the evolution from their initial Twitter search tool to what is now known as Perplexity. He highlights the vision of creating a more efficient search experience across various databases and the potential value of advanced search capabilities, emphasizing the need for innovative solutions in the search landscape.
"insane and they showed it to a bunch of people and they all allowed it main there are two reasons one something like that never existed before like uh you could never actually goarch exactly to this d..."
In this segment, Aravind shares insights on their strategy of using a simpler, heuristic approach to AI that capitalizes on the advancements in model capabilities. He discusses how this method improved user experience by reducing latency and enhancing the speed of responses.
"and uh you also stand a chance against the Legacy system that Google has built which is a lot more in the first style so we thought okay we would try to build a more General solution and then we proto..."
Aravind recounts the moment Perplexity gained traction on social media, including a humorous incident involving a user searching for their own biography. This sparked discussions about trust in AI-generated answers and led to increased engagement with the platform.
"used to uh like like links appearing instantly on a traditional search right even then by the way the first version we launched which is the answer version uh took 7 seconds or something to um because..."
Aravind reflects on his initial thoughts about competing with Google, emphasizing the unique challenges posed by Google's cluttered search interface. He discusses the competitive landscape, including concerns about Microsoft and the strategic decisions that shaped Perplexity's direction.
"exponentially so I was like okay this is there's something here it's not worth killing and pivoting to Enterprise you have this like initial momentum and you you said earlier it wasn't until hindsight..."
In this closing segment, Aravind emphasizes the importance of user feedback in product development. He shares insights from Larry Page about the philosophy that 'the user is never wrong,' highlighting how this principle guides their approach to building Perplexity.
"else here yeah having spent almost a decade at Google myself um I see a lot of the culture of the early days of Google like the things I've learned about Larry or about SAR and I see a lot of that in ..."
In this segment, Aravind reflects on the design principles he learned from his time at Google, focusing on creating intuitive user experiences. He discusses features like auto-suggest and spell-check, emphasizing the need for products to anticipate user needs and streamline interactions.
"other way of Designing products is like make the user be a better prompt engineer mhm blame the user and tell them to be a better prompt engineer teach them educate them get them to do it the way that..."
Aravind explains how Perplexity measures success through daily query metrics, drawing parallels to Google's early strategies. He highlights the challenges of user retention and the importance of ensuring that the product meets user needs effectively.
"don't need to take your mouse and place it there it sounds like your your main metric that you care about is number of queries per day which is exactly what Google did I think in the early days right ..."
Aravind shares insights into how he manages his growing team at Perplexity, emphasizing a culture of transparency and direct communication. He discusses the importance of maintaining high standards for product quality and fostering an environment where team members feel empowered to address issues.
"look every week see the weekly growth rates um see the monthly growth rates and like if something declined then discuss about it figure out ways to actually freak out if something declines we do and s..."
In this segment, Aravind talks about the value of receiving honest feedback from users, particularly on social media. He contrasts the candidness of online interactions with the politeness often found in emails, stressing that brutal honesty can reveal critical bugs and areas for improvement.
"select for that when you were hiring like people who were very product Centric and in the details I wouldn't say I I explicitly had that as a Criterion but I look for people who cared about doing good..."
Aravind discusses the challenges of scaling a startup without losing agility. He reflects on the growing pains of Perplexity, including the need for rigorous testing and the balance between speed and quality as the company expands its operations.
"the worst bugs and uh things that people are afraid to say oh and in person is the worst where you go show someone something and they're just going to tell you good things even if they hate it I kind ..."
Aravind shares his vision for the future of Perplexity, aiming to position it as a more intelligent alternative to Google. He discusses upcoming features and the goal of providing users with highly relevant answers for specific queries, such as shopping and travel.
"it yeah okay let's talk a little bit about the future um you know I've seen you you're most recent launches are kind of like in different directions more verticalized or more specific around shopping ..."
Aravind reflects on the competitive landscape of search technology, particularly in relation to giants like Google. He emphasizes the importance of user obsession and product taste as key differentiators for Perplexity. Aravind argues that while Google has extensive resources, Perplexity's focus on user experience and innovative solutions positions it well to tackle the challenges of modern search and information retrieval.
"MH you just have to like persever and work for a decade or two on this problem if I talk to people at Google they would say yep that's what we're building in fact I know they've been saying that for m..."