Building Search for AI Agents with Exa CEO Will Bryk
Exa CEO Will Bryk discusses how his company is building a search engine specifically designed for AI agents rather than human consumers, arguing that agentic search requires fundamentally different architecture than Google's click-optimized system. He explains why LLMs have made it possible for a small team to compete with Google, and predicts agentic search will surpass Google Search in revenue by the 2030s.
Summary
Will Bryk, co-founder and CEO of Exa, speaks with a16z's Sarah Wang about the origins of Exa and his lifelong mission to build perfect search. Bryk describes how his interest in information quality began in childhood and evolved through writing a history book, where he realized Google's surface-level results were inadequate for deep research. He and co-founder Jeff built a crowdsourced search engine in college, then founded Exa in 2021 when Transformer models made it technically feasible to beat Google using neural retrieval systems.
Bryk distinguishes between human search and agentic search, arguing they require fundamentally different systems. While Google is optimized for billions of tired consumers typing imprecise keyword queries, AI agents need comprehensive results (thousands to tens of thousands, not just ten), complex semantic queries, fine-grained controllability, and lower latency. He uses the analogy of building a search engine for sloths versus humans — agents are so different from humans that a human-optimized system simply doesn't serve them well.
On competition with Google, Bryk acknowledges Google excels at consumer use cases but argues it fails at deep, complex queries like finding every competitor in a market, recruiting candidates, or understanding a historical period in depth. He notes that Google's massive click-data advantage is largely irrelevant for agent use cases, and that LLMs have democratized capabilities like re-ranking that previously required hundreds of engineers, enabling Exa's small team to build a superior product for specific use cases.
Bryk discusses the 'tokenpocalypse' — the problem of runaway AI inference costs — and argues that retrieval-augmented systems using smaller models are the solution. He claims Exa can save customers up to 20x on costs by extracting only the most relevant information and enabling smaller models to perform like larger ones. He predicts this shift toward smaller, tool-using models will be clearly visible by end of 2026.
Bryk frames many major societal problems as search problems in disguise: political polarization (people receiving inaccurate information), loneliness (inability to find people with shared interests), and knowledge gaps. He also discusses Exa's go-to-market intelligence product, its use by coding agents like Devin, and research directions including RL on search tools, where Exa outperformed Google-wrapped search in efficiency and quality.
On company culture, Bryk emphasizes passion ('fire in the eye') as the primary hiring criterion, citing the meritocratic opportunity enabled by agentic tools. He describes an environment where engineers work on whatever excites them most, laugh loudly at 8pm, and are unified by the mission of perfect search. He draws leadership inspiration from his SpaceX internship and Elon Musk's detail-orientation and memetic communication style.
About this episode
Sarah Wang speaks with Exa cofounder and CEO Will Bryk about building search infrastructure for the AI era. The conversation covers Exa’s origins, why traditional search engines were not designed for AI agents, and how search changes when the user is no longer a human but an autonomous system. They discuss retrieval, agent workflows, coding agents, data access, and why search may become a foundational layer for the emerging agent economy. Along the way, Bryk shares his views on AI-native products, the future of information discovery, and why some of the most important problems in technology can ultimately be framed as search problems.
Key Insights
- Bryk argues that Google's massive human click-data advantage is largely irrelevant for AI agent use cases, since agents don't benefit from knowing what humans historically clicked on — leveling the playing field for Exa.
- Bryk claims LLMs have collapsed the engineering complexity of search components like re-ranking, which previously required hundreds of engineers at Google but can now be handled by a couple of people using trained models, enabling Exa's sub-100-person team to outperform Google in specific domains.
- Bryk contends that a large portion of knowledge work is actually a search problem rather than an intelligence problem, meaning the bottleneck for most enterprise AI tasks is retrieval quality, not model capability.
- Bryk predicts agentic search will generate more revenue than Google Search by the early 2030s, based on projections that agents will perform millions of searches per user per day versus the handful humans perform, creating an infrastructure demand comparable to electricity.
- Bryk argues that retrieval-augmented small models can solve the 'tokenpocalypse,' claiming Exa can save customers up to 20x on inference costs by extracting only the most relevant document content for agents to consume.
- Bryk frames political polarization and loneliness as fundamentally search problems — people receive misleading information and fail to find compatible people not due to lack of desire but due to inadequate search infrastructure.
- Bryk found through RL experiments that agents using Exa as their search tool made fewer total search calls and achieved higher task performance compared to agents using Google-wrapped search, because Exa accepts complex semantic queries without forcing compression into keywords.
- Bryk identifies infrastructure scalability — not model intelligence or data — as the near-term bottleneck for agentic search, noting that handling 100-1000x more queries than Google requires vector database and retrieval infrastructure that hasn't been built yet.
Topics
Transcript
Search is the gateway to the world's information. If you can make it perfect, then that has so many downstream positive implications for the world. You can kind of think of Google as being synonymous with search, right? It's one of the greatest technopolies of the last few decades. If you want to go really deep into some topic, Google fails. Most people want to understand the world, but they're getting fed information that's just like misleading in some way or straight up wrong. And if everyone had information that was accurate, most reasonable people would be reasonable. We have a family open club, Michael Claudeberg, and we wanted to give it web access. And he was like, I recommend…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The a16z Show
The Top 100 Consumer AI Apps: Who’s Actually Paying?
A16Z's seventh edition Top 100 Consumer AI Apps Report reveals that while half of Americans use AI, only 4.5% pay for subscriptions, with spending heavily concentrated among developers and power users. The report highlights emerging personal agents as a major trend, significant divergence between ChatGPT, Claude, and Gemini, and substantial untapped opportunities in categories like dating, social, shopping, and entertainment.
David George & Jack Altman on AI, Autonomy, and the Next $25 Trillion
David George from Andreessen Horowitz argues that AI, autonomy, and robotics will create over $25 trillion in market value over the next decade, with success across multiple layers of the stack—frontier models, open source, and applications will all flourish due to massive supply constraints and early-stage demand penetration among knowledge workers.
Beyond the God Model | Alex Atallah & Amjad Masad
Alex Atallah (OpenRouter) and Amjad Masad (Replit) discuss how the future of AI development lies in specialized, diverse models working together rather than relying on single AGI-like models. They argue that enterprises need independence from foundation model providers through model diversity, cost efficiency, and ownership of their own AI capabilities, while also addressing the risks of general-purpose agents and the importance of model specialization.
Why AI Agents Can Beat the Incumbents
Leo, an AI-native startup, is building multi-agent systems to automate end-to-end procurement processes for enterprises by coordinating work across multiple departments, stakeholders, and external systems that traditional procurement software cannot handle. The company differentiates itself from incumbents by owning the entire job workflow rather than being confined to a single system of record, and builds customer trust through human-in-the-loop approaches before scaling to fully autonomous negotiations.
The $1 Trillion AI Buildout | State of Markets
A16Z partners discuss the $1 trillion AI infrastructure buildout, arguing it's driven by real earnings growth rather than inflated valuations, with adoption still extremely early at the enterprise level. They highlight opportunities across consumer agents, robotics, autonomous vehicles, and enterprise diffusion, while noting that the market's 90% gain since ChatGPT reflects fundamental business performance rather than speculative excess.