How did Perplexity start, and what was its initial product?

Perplexity began without a clear product idea but focused on building natural language SQL for searching relational databases, specifically Twitter data. This evolved into a broader web search tool that provides answers with cited sources, distinguishing itself from platforms that might not prioritize accuracy.

Why did Perplexity choose to build its own browser instead of relying on existing search engines?

The founders realized that dominant players like Google have conflicting incentives (ads vs. direct answers) and that building a proprietary browser (Comet) allows for a more integrated AI assistant experience, control over data, and the ability to perform complex agentic tasks without OS restrictions.

How does Perplexity aim to achieve accuracy and avoid hallucinations in its AI-generated answers?

Perplexity is actively building internal benchmarks and improving its search index to capture better web page snippets. They also leverage large language models for multi-step reasoning on queries, aiming to reduce costs and enhance accuracy to minimize incorrect data.

What is Perplexity's business model, and how does it differ from Google's?

Perplexity focuses on subscription revenue and usage-based pricing for agent tasks, avoiding an ad-based model like Google's. They believe this aligns better with providing accurate, unbiased answers and avoids the conflict of interest inherent in advertising.

What advice does Aravind Srinivas have for aspiring entrepreneurs?

Srinivas advises entrepreneurs to work incredibly hard, embrace the fear of big companies copying their ideas, and focus on building a strong brand identity through speed and user focus. He also draws inspiration from Elon Musk's persistence in the face of repeated failures.

How does Perplexity plan to build network effects and create a sticky product?

While brand recognition is a network effect, Perplexity is betting on its browser to create deeper product stickiness. By integrating browsing history, passwords, wallet, and agent tasks, they aim to build a more indispensable tool that users are less likely to switch away from.

What is Perplexity's perspective on the future of the web in an era of AI-driven search impacting website traffic?

Perplexity anticipates a more skewed web where well-known brands with direct organic visits will thrive, while those gaming SEO will struggle. They believe their AI search model, while potentially reducing traffic to some sites, offers a more efficient way to access information.

Key Moments

Aravind Srinivas: Perplexity's Race to Build Agentic Search

Y Combinator

Science & Technology6 min read44 min video

Jul 11, 2025|128,023 views|2,208|107

YC Y Combinator

Save to Pod

Key Moments

On this page

TL;DR

Perplexity is betting its future on a browser, aiming to be an 'assistant' rather than just a search engine, but faces stiff competition from tech giants like Google and OpenAI.

Key Insights

Perplexity experiences daily infrastructure issues due to high user growth, necessitating a rebuild for 10x scaling.

The company's 'big bet' is on a browser, referred to as 'Comet,' designed to function as a cognitive operating system with integrated agents.

Aravind Srinivas actively participates in bug triage and fixes himself, reflecting a hands-on leadership style aimed at rapid innovation.

Perplexity launched shortly after ChatGPT in late 2022, capitalizing on the then-lack of web search in ChatGPT and the rapid evolution of AI models.

Google's business model, reliant on ads, inherently conflicts with providing users with direct, unbiased answers, creating an opening for Perplexity.

Partnerships with companies like Selfbook, TripAdvisor, and Shopify are crucial for Perplexity's agentic capabilities and business model.

Scaling challenges and the grand vision of a cognitive browser

Perplexity is experiencing significant daily infrastructure challenges due to its rapid user growth, indicating a need to scale its systems for a 10x increase in capacity. This growth, however, fuels the company's major strategic shift towards developing a browser, internally codenamed 'Comet.' This new browser is envisioned not just as a tool for navigation and informational queries, but as an 'assistant' and a 'cognitive operating system.' It will integrate agents capable of performing complex, asynchronous tasks, pulling data from personal contacts, emails, calendars, and social media to execute research on real estate, markets, and more. The goal is to make the browser a cloud-like environment where multiple parallel processes run autonomously, transforming the user's new tab page and sidebar into an active AI assistant. This ambitious undertaking is framed as the future of the company, moving beyond being just another search or chatbot application.

The intense competitive landscape and the imperative of speed

Perplexity operates in a hyper-competitive environment, with major players like Google and OpenAI also heavily investing in AI-powered search and browsing experiences. Both companies have introduced features like 'AI Overview' and 'AI Mode,' which have been perceived by some as direct competitive threats. Aravind Srinivas acknowledges this, stating that if a sector is profitable, well-funded competitors will inevitably enter. However, he emphasizes that Perplexity's unique advantage lies in its singular focus on accuracy and task orchestration, particularly within the browser domain, which is harder to replicate than a simple chatbot. The prevailing strategy to counter these formidable competitors is speed: innovating and moving faster than everyone else. Srinivas likens this to running a marathon at an extremely high velocity, highlighting the relentless pace required to stay ahead.

Leadership in action: The CEO's role in debugging and innovation

Aravind Srinivas's leadership style is characterized by a deep involvement in the product's technical details, exemplified by his personal dedication to triaging and fixing bugs, even stopping all other activities to address them. This hands-on approach contrasts with the typical delegation practices of CEOs in larger corporations. Srinivas views this not as trivial but as essential for understanding and improving the product. He notes that even leaders at companies like Google are now engaging in bug support, suggesting a potential trickle-down effect towards more direct involvement. This commitment to detail and problem-solving is presented as a core element of Perplexity's culture, enabling the company to innovate rapidly and maintain a competitive edge, especially when facing large, established players.

Founding Perplexity: From niche search to a web-wide ambition

Perplexity's origin story deviates from the conventional advice of starting with a fully formed idea. Instead, the founders began by building tools for specific problems, such as a natural language-to-SQL search engine for relational databases like Twitter. This initial focus on efficient searching over structured data, inspired by early concepts like Facebook's graph search, demonstrated sustained user interest. Recognizing the limitations of structuring all web data, the team pivoted to leverage language models for reasoning and parsing unstructured data across the entire web. This led to the development of Perplexity as a general-purpose AI-powered search engine that provides direct answers with cited sources. The company launched its Discord bot version shortly after ChatGPT, strategically positioning itself at a time when competing AI tools lacked robust web search capabilities.

The innovator's dilemma and the fundamental conflict with Google's model

A pivotal moment for Perplexity was Google's announcement of Bard. Srinivas identified a fundamental conflict in business models: Google's reliance on advertising makes it inherently disincentivized to provide direct answers that bypass traditional ad platforms like booking sites or e-commerce retailers. Unlike Google, which would face a conflict of interest, Perplexity can build a business model around user subscriptions and transaction fees without compromising its core function of delivering accurate, sourced answers. This 'innovator's dilemma' also extends to AI model development; startups like Perplexity, by leveraging external foundation models, could initially possess superior AI capabilities than internal Google models. Furthermore, the higher risk tolerance of startups—where a single mistake doesn't tank stock prices, unlike Google—allows for faster iteration and learning.

The browser as the next frontier for AI agents

Perplexity views the browser as the critical next step beyond chatbots and search. The browser, exemplified by their focus on 'Comet,' can act as an abstraction layer for AI agents, overcoming limitations of current chatbots. Unlike chatbots, browsers can interact with other tabs, access browsing history, fill forms, make purchases, and perform recurring tasks autonomously. Srinivas believes building truly capable mobile agents will be significantly challenging without a browser OS that can circumvent app store restrictions and integrate with various third-party applications. This approach offers greater control and flexibility compared to relying solely on MCP servers or specific app integrations, positioning the browser as the foundational platform for advanced AI assistance. The transition from search to browser, akin to Google's evolution from search to Chrome, is seen as essential for capturing a massive user base and daily queries.

Building sticky products and exploring new business models

Perplexity initially focused on building brand and narrative around accuracy and speed, which provided some resilience against larger competitors. However, the company recognizes the need for deeper user engagement. They believe the browser, by integrating personal data, browsing history, passwords, and wallet information, alongside running persistent agent tasks, can create significant stickiness and network effects. Furthermore, Perplexity is actively exploring new business models beyond subscriptions. While acknowledging they may not reach Google's profit scale, they aim for substantial revenue through subscriptions and usage-based pricing for agent tasks. They also see potential in taking a cut from transactions facilitated by AI. This diversified approach seeks to build a sustainable business that isn't solely reliant on advertising, unlike their primary competitors.

Navigating the evolving web and maintaining content integrity

The rise of AI search engines like Perplexity raises questions about the future of the web and content creation, as direct answers may reduce traffic to original websites. Srinivas acknowledges that the web will likely become more skewed, favoring well-known brands that can maintain direct organic visits, while SEO-focused sites may struggle. Regarding IP and bias, Perplexity emphasizes sourcing data from reputable providers and presenting multiple perspectives on subjective topics, aiming for objective truth where possible (e.g., sports scores, weather). For news and opinion-based content, they strive to surface diverse viewpoints rather than taking a definitive stance. Addressing concerns about hallucinations, they focus on improving search indexing, capturing better snippets, and using faster models for multi-step reasoning, while also acknowledging the reliance on trust built over time through consistent accuracy.

Mentioned in This Episode

●Software & Apps

●Companies

●Organizations

●People Referenced

Common Questions

Perplexity's primary strategy is to focus on building a superior browser experience integrated with AI and agentic capabilities, which they believe is harder to replicate than a simple chat tool. They aim for speed and accuracy, differentiating themselves from competitors exploring similar AI-driven search functionalities.

Topics

AI & Machine Learning Technology & Innovation Product Development Future Of The Internet LLM Applications Startup Strategies Competitive Analysis AI Search Engines Browser Technology

Mentioned in this video

People

Sundar Pichai

CEO of Google and Alphabet. His potential view on Perplexity's direction and his actions regarding bug support are discussed.

Sam Altman

CEO of OpenAI. Mentioned in the context of how he might view Perplexity's browser strategy.

Elon Musk

Founder of SpaceX, Tesla, and owner of X. His resilience in the face of failure, particularly from YouTube videos, is cited as inspiration for Perplexity's CEO.

Nikil Kamat

A person who apparently interviewed with Perplexity for an internship, as referenced by a questioner.

Companies

Shopify

An e-commerce platform. It is listed as a partner of Perplexity, indicating integrations for shopping functionalities.

Yelp

A platform for local business reviews. Perplexity works with Yelp for map collaborations and surfacing information.

FMP Sports

A data provider in the sports vertical that partners with Perplexity.

Samsung

A multinational electronics corporation. Talks about Perplexity potentially being pre-installed on Samsung phones are mentioned.

Y Combinator

A startup accelerator that provides funding and mentorship to early-stage companies. The speaker mentions applying to Y Combinator and its advice for starting companies.

OpenAI

An artificial intelligence research lab that developed ChatGPT. Their competitive moves and well-funded status are discussed in relation to Perplexity.

Selfbook

A company that powers hotel bookings on Perplexity, enabling native purchasing experiences.

Bloomberg

A financial media and software company. Cited as a source for Perplexity's potential valuation increase.

Anthropic

An AI safety and research company. Mentioned as a competitor developing their own tools, similar to Perplexity.

Twitter

A social media platform. Perplexity's initial focus was on improving search capabilities for Twitter data.

TripAdvisor

A travel guidance company. They collaborate with Perplexity to surface reviews for hotels and other places.

Costco

A membership-based warehouse club. Mentioned as an example of a business with access to a user base that might not be heavily engaged with AI, indicating a potential target audience for go-to-market strategies.

Stats Perform

A sports data and analytics provider that collaborates with Perplexity.

NVIDIA

A technology company specializing in graphics processing units and AI. Mentioned in the context of partnerships for shipping AI models.

Software & Apps

Comet

The name of Perplexity's new browser project, envisioned as a cognitive operating system and an assistant.

Cursor

An AI-first code editor. Perplexity uses it internally, and its success is noted as an example of a strong product attracting acquisition interest.

Chrome

Google's web browser. The browser's role in Google's success and Perplexity's ambition to create a similar platform with its own browser (Comet) is discussed.

Bard

Google's conversational AI service. Its announcement by Sundar Pichai marked a key moment when Perplexity's founders realized they needed to build a separate product.

Claude

An AI assistant possibly developed by Anthropic. Mentioned as a tool that seems 'far smarter' than Cursor.

Perplexity

An AI-powered search engine that provides direct answers with citations. It is the main subject of the discussion, with its growth, product strategy, and competitive landscape being explored.

ChatGPT

A large language model developed by OpenAI, known for its conversational AI capabilities. Perplexity's launch strategy and competition with ChatGPT are discussed.

Gemini

Google's multimodal AI model. Mentioned as one of several AI assistants users might consult.

Facebook Graph Search

An earlier search feature on Facebook that allowed users to search within their social graph. This inspired Perplexity's initial vision for searching structured data.

GitHub Copilot

An AI pair programmer developed by GitHub. It's used by Perplexity engineers alongside Cursor for coding tasks.

SwiftUI

A declarative UI framework for developing apps across Apple platforms. Used in Perplexity's front-end development, with Cursor assisting in code modifications.

Organizations

Wikipedia

A free online encyclopedia. Mentioned as a source for objective information, but its limitations in capturing diverse perspectives on subjective topics are discussed.