Field notes on agents, computers, and the work they do.

APR 11 · 2026 / Philosophy

Do We Really Need to Build AI Agents for Data?

The software for data work already exists and is mature. The only problem was UX. AI coding agents fix that — no specialized data agent required.

APR 1 · 2026 / Engineering

How We Built Sub-Second Sandbox Startup

A deep dive into our two-tier snapshot architecture that delivers 400ms hot starts and permanent filesystem persistence — with zero data loss.

MAR 28 · 2026 / Philosophy

Skills Need a Browser

Before the web, information was siloed on individual machines. Hyperlinks and browsers changed everything. AI skills are at the same inflection point — the capabilities exist, but there is no execution layer for ordinary people. We are building it.

MAR 27 · 2026 / Philosophy

Rebyte Is Not a Super Agent

The market is full of enterprise agent systems — vertical solutions and generic "do everything" platforms. Rebyte takes a different path: four fundamental primitives that anyone can understand and combine.

MAR 15 · 2026 / Engineering

From Serverless Functions to Serverless Agents

AWS Lambda changed backend development by eliminating idle server costs. AI agents need the same revolution — but with full Linux environments, persistent state, and unlimited runtime. This is what Lambda for agents looks like.

MAR 4 · 2026 / Engineering

Why Every Agent Needs a 'Box'

AI agents are evolving from local geek toys into enterprise-grade productivity tools. The key infrastructure enabling this transformation is a new kind of sandbox — fast startup, long-running, persistent, on-demand, self-identifying, and network-isolated.

MAR 3 · 2026 / Engineering