What are the key takeaways from this a16z Podcast episode?

Key insights include: **Agentic vs. Human Search Architecture:** Agents require fundamentally different search infrastructure than humans — they need thousands to 10,000 results per query rather than 10, tolerate variable latency, and submit complex semantic queries without keyword compression. Building for agents means exposing granular toggles like domain filters, keyword controls, and semantic switches that human-facing search engines deliberately hide.; **Token Efficiency via Retrieval:** Pairing smaller language models with high-quality retrieval cuts inference costs by up to 20x compared to running large models alone. Exa extracts only the most relevant document segments before passing content to models, dramatically reducing input token consumption. The practical architecture: a large model orchestrates tasks, small models execute them using retrieval to compensate for reduced parameter counts.; **Google's Click Data Advantage Doesn't Transfer:** Two decades of human click-signal data — Google's core ranking moat — provides minimal advantage when serving AI agents. Agents don't click, don't browse casually, and don't benefit from popularity-weighted results. This levels the competitive playing field, allowing a sub-100-person team to build retrieval quality that outperforms Google on deep, complex, business-critical queries.

What did Exa Ceo Will Bryk discuss on a16z Podcast?

Exa CEO Will Bryk explains how his company builds search infrastructure specifically for AI agents, which require deeper context, comprehensive results, and complex query handling that Google's consumer-oriented, click-data-driven architecture was never designed to deliver, positioning agentic search to surpass Google Ads revenue by the 2030s. Key topics include: **Agentic vs. Human Search Architecture:** Agents require fundamentally different search infrastructure than humans — they need thousands to 10,000 results per query rather than 10, tolerate variable latency, and submit complex semantic queries without keyword compression. Building for agents means exposing granular toggles like domain filters, keyword controls, and semantic switches that human-facing search engines deliberately hide.; **Token Efficiency via Retrieval:** Pairing smaller language models with high-quality retrieval cuts inference costs by up to 20x compared to running large models alone. Exa extracts only the most relevant document segments before passing content to models, dramatically reducing input token consumption. The practical architecture: a large model orchestrates tasks, small models execute them using retrieval to compensate for reduced parameter counts..

How long is this episode of a16z Podcast?

This episode is 49 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

a16z Podcast

Building Search for AI Agents with Exa CEO Will Bryk

June 6, 2026

49 min episode · 2 min read

Exa Ceo Will Bryk

Episode

49 min

Read time

2 min

Topics

Productivity, Leadership, Design & UX

AI-Generated Summary

Published Jun 6, 2026

Key Takeaways

✓Agentic vs. Human Search Architecture: Agents require fundamentally different search infrastructure than humans — they need thousands to 10,000 results per query rather than 10, tolerate variable latency, and submit complex semantic queries without keyword compression. Building for agents means exposing granular toggles like domain filters, keyword controls, and semantic switches that human-facing search engines deliberately hide.
✓Token Efficiency via Retrieval: Pairing smaller language models with high-quality retrieval cuts inference costs by up to 20x compared to running large models alone. Exa extracts only the most relevant document segments before passing content to models, dramatically reducing input token consumption. The practical architecture: a large model orchestrates tasks, small models execute them using retrieval to compensate for reduced parameter counts.
✓Google's Click Data Advantage Doesn't Transfer: Two decades of human click-signal data — Google's core ranking moat — provides minimal advantage when serving AI agents. Agents don't click, don't browse casually, and don't benefit from popularity-weighted results. This levels the competitive playing field, allowing a sub-100-person team to build retrieval quality that outperforms Google on deep, complex, business-critical queries.
✓RL Training on Search Tools Yields Measurable Gains: Exa's research applying reinforcement learning directly to search tool selection — comparing Google SERP wrapping against Exa — showed that agents trained on Exa made fewer total search calls while achieving higher task performance. The mechanism: Exa's architecture accepts complex natural-language queries, so agents don't waste calls reformulating needs into keyword approximations.
✓Go-to-Market and Recruiting as Unsolved Search Problems: Company and people search remains a genuinely unsolved problem — no current tool reliably surfaces every competitor across global markets or every qualified candidate matching specific criteria. Exa is building go-to-market intelligence products targeting this gap, using it internally as a live testbed, treating comprehensive entity retrieval as the core technical challenge rather than a UI or workflow problem.

What It Covers

Exa CEO Will Bryk explains how his company builds search infrastructure specifically for AI agents, which require deeper context, comprehensive results, and complex query handling that Google's consumer-oriented, click-data-driven architecture was never designed to deliver, positioning agentic search to surpass Google Ads revenue by the 2030s.

Key Questions Answered

•Agentic vs. Human Search Architecture: Agents require fundamentally different search infrastructure than humans — they need thousands to 10,000 results per query rather than 10, tolerate variable latency, and submit complex semantic queries without keyword compression. Building for agents means exposing granular toggles like domain filters, keyword controls, and semantic switches that human-facing search engines deliberately hide.
•Token Efficiency via Retrieval: Pairing smaller language models with high-quality retrieval cuts inference costs by up to 20x compared to running large models alone. Exa extracts only the most relevant document segments before passing content to models, dramatically reducing input token consumption. The practical architecture: a large model orchestrates tasks, small models execute them using retrieval to compensate for reduced parameter counts.
•Google's Click Data Advantage Doesn't Transfer: Two decades of human click-signal data — Google's core ranking moat — provides minimal advantage when serving AI agents. Agents don't click, don't browse casually, and don't benefit from popularity-weighted results. This levels the competitive playing field, allowing a sub-100-person team to build retrieval quality that outperforms Google on deep, complex, business-critical queries.
•RL Training on Search Tools Yields Measurable Gains: Exa's research applying reinforcement learning directly to search tool selection — comparing Google SERP wrapping against Exa — showed that agents trained on Exa made fewer total search calls while achieving higher task performance. The mechanism: Exa's architecture accepts complex natural-language queries, so agents don't waste calls reformulating needs into keyword approximations.
•Go-to-Market and Recruiting as Unsolved Search Problems: Company and people search remains a genuinely unsolved problem — no current tool reliably surfaces every competitor across global markets or every qualified candidate matching specific criteria. Exa is building go-to-market intelligence products targeting this gap, using it internally as a live testbed, treating comprehensive entity retrieval as the core technical challenge rather than a UI or workflow problem.

Notable Moment

Bryk argues that political polarization is fundamentally a search problem — most people want accurate information but receive misleading or incomplete content, causing otherwise reasonable people to hold unreasonable views. He includes himself, acknowledging his own beliefs are likely distorted by imperfect information access.

Know someone who'd find this useful?

Episode Transcript

Search is the gateway to the world of 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 technopies of the last few decades. If you wanna go really deep into some topic, Google fails. Most people wanna understand the world, but they're getting fed information that's just, like, you know, misleading in some way or straight up wrong. And if everyone had, like, information that was accurate, most reasonable people would be reasonable. We have a family open claw, Michael Clodberg, and we wanted to give it web access, and he was like, I recommend Exa. World of agents searching is just completely different from human searching. An agent doesn't just want 10 pieces of information. It wants everything. Because that's, like, you could search something and then get not just, like, 10 results or a 100 results, but a thousand results or 10,000. How have we, a team that, you know, has always been below a 100 people, been able to build a search engine that's better than Google in all sorts of ways? Well, it's because For most of the Internet era, search was built for humans. But AI agents search differently. They need deeper context, more complete information, and the ability to navigate far more complex questions than a traditional search box was designed to answer. That shift is creating an entirely new set of challenges around retrieval, knowledge discovery, and how information is organized online. Sarah Wang speaks with Exa cofounder and CEO Will Brick about search, AI agents, and the future of information retrieval. Welcome, Will. Thank you for being here. Well, excited to be here. So I wanna start with the origin story. You've been interested in search for a long time. In fact, you and your cofounder, Jeff, actually started building a mini search engine in college, which is not what I was doing in college. Can you say more about when you started getting interested in search and why you wanted to solve this problem? Yeah. Yeah. Sure. So I would say it's a life mission. So since I was a kid, I've cared about finding the highest quality knowledge. Right? I was obsessed. And then in high school, I wanted to start a new type of news organization because I thought we're a civilization that got to the moon, we split the atom, and yet we can't understand what's going on at the border or, so in science news, like, why can't we fully understand any topic? And then in college, you know, I was roommates with Jeff, and we were like, we could just build a better search using crowdsourcing the highest quality links. And we did build a pretty solid search. But then five years ago, so in 2021, that's when Transformers started to get really good. And it suddenly became possible …

Get the full transcript (11,368 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all a16z Podcast transcripts →