introducing kernel

I'm excited to (finally!) share what Rafael Garcia and I have been working on for the past several months. We're calling it Kernel.

In February, I left Cash App to start a new company. The pace of AI innovation was—and still is—dizzying. It was clear to me that if I wanted to be part of this next wave, I needed to be on the frontlines. What was especially obvious to Raf and me from the start: new capabilities emerge every week, but developer tooling and infrastructure isn't keeping up.

Paradigm shifts are incredible moments to build, and some of the most interesting companies emerge when you follow the second- and third-order effects of new technologies: think Uber, Instagram, and Waze from the mobile era.

If you believe autonomous agents will become embedded in every corner of our lives—handling customer support, scheduling, research, data analysis, and beyond— then they'll need access to the same tools and interfaces we do. APIs and MCP servers are great when they're available, but 95% of the internet only exists as websites. Most things we use in our day-to-day lives still live behind logins, forms, dashboards, and buttons. Agents need to click them.

Browser automation isn't new, but the rise of agentic, autonomous reasoning is accelerating demand for programmable infrastructure that can interact with the internet. Developers more than ever need these rails to work fast at scale. Kernel exists to solve these problems.

In April, we released our first open-source project: we put a Chrome browser on a unikernel that boots in milliseconds. We weren't sure if anyone would care, but the response was immediate. We made it to the front page of Hacker News, and hundreds of developers starred our repo. Our small Discord community 10x'ed. We were off to the races!

Kernel provides high-performance, production-grade browser infrastructure for web automations and AI agents, providing fully isolated browser sessions for AI agents in under 20ms. It's built for developers who want to automate the web, whether you're running an agent, orchestrating a workflow, QA'ing an app, or doing deep research. Under the hood, we've built an infrastructure platform supporting sub-100 millisecond cold starts, session persistence, and first class browser support, enabling developers to launch their web-based AI agents in seconds.

We're currently in private beta. If you're building web automations for your start-up or developing a new web agent framework, I'd love to hear from you.

And most importantly: we're hiring! Message me if you're excited about low-latency systems, developer tools, and making the web programmable. Join our Discord to follow along.

may 29, 2025
← back