Harnessing the Web with Agentic Tools With CEOs of Exa and Browserbase
People
-
Will Bryk Guest
-
Paul Klein Guest
-
Madison McIlwain Guest
Watch on
Agents are about to run far more searches than humans do, and most of the web they need to use was built for people. Exa CEO Will Bryk and Browserbase CEO Paul Klein explain how they are building the search and browser infrastructure that lets AI agents find information and take action. In this IA40 Summit panel, moderated by Vercel’s Madison McIlwain, Will (founder of Exa, the search engine built for AI agents) and Paul (founder of Browserbase, the headless browser infrastructure company) talk about what is breaking as agents use the web at scale and how their companies are designed around it. What you’ll take away: + Why Will expects agent searches to pass Google’s roughly 15 billion daily queries by early next year, and reach about 1,000 times more in a few years + Why Paul compares browser agents to Waymo and expects AI to use the web as it exists today + How computer-use models passing the human score on OS World Verified in 2026 changed what browser agents can do + How both companies stay model-agnostic and why both are bullish on model routers + How Exa and Browserbase find compute when it is hard to buy, including a distributed network and a data center purchase Chapters: 00:00 – How the Exa and Browserbase Founders Ended Up Building Infrastructure for AI Agents 03:09 – Why Exa Set Out to Redesign Information Tools for an AI World 05:52 – Why AI Will Use the Web We Already Have, With Waymo as the Playbook 07:05 – Why Agent Search Volume Will Require Infrastructure Bigger Than Google's 09:42 – How Computer-Use Models Passed the Human Score and Changed Browser Agents 12:17 – Why Infrastructure Startups Should Stay Model-Agnostic as Labs Ship New Models 16:07 – Why Giving Agents Better Tools Beats Giving Them Smarter Models 17:15 – How Legacy Shipping Port Software Makes the Case for Browser Agents Over APIs 18:35 – How Exa Builds a Distributed Network of Compute When Hardware Is Sold Out 21:12 – Why Companies Need an Agent That Works as One Company-Wide Brain 22:57 – How Model Routing Creates a Data Flywheel for Browserbase 24:50 – Why Exa's Will Bryk Says His P(doom) Is Low
Chapters
- 0:00 How the Exa and Browserbase Founders Ended Up Building Infrastructure for AI...
- 3:09 Why Exa Set Out to Redesign Information Tools for an AI World
- 5:52 Why AI Will Use the Web We Already Have, With Waymo as the Playbook
- 7:05 Why Agent Search Volume Will Require Infrastructure Bigger Than Google's
- 9:42 How Computer-Use Models Passed the Human Score and Changed Browser Agents
- 12:17 Why Infrastructure Startups Should Stay Model-Agnostic as Labs Ship New Models
- 16:07 Why Giving Agents Better Tools Beats Giving Them Smarter Models
- 17:15 How Legacy Shipping Port Software Makes the Case for Browser Agents Over APIs
- 18:35 How Exa Builds a Distributed Network of Compute When Hardware Is Sold Out
- 21:12 Why Companies Need an Agent That Works as One Company-Wide Brain
- 22:57 How Model Routing Creates a Data Flywheel for Browserbase
- 24:50 Why Exa's Will Bryk Says His P(doom) Is Low