The Agent Cloud: Databricks’ Bet on the Future of AI — Matei Zaharia and Reynold Xin
- Source
- youtube.com
- Author
- Latent Space
- Date
Context
Matei Zaharia and Reynold Xin of Databricks explain in a Latent Space interview that their open-source Omnigent came out of two lines of work: engineers wrapping coding agents like Claude Code and Codex in their own workflows, and teams building data agents that needed to swap models and software every few months. They say a useful agent also needs shared sessions, history and search, and security controls if colleagues are to use it.
Xin adds a personal trigger: during a week of nonstop coding with agents he kept his laptop tethered to his phone in a car so sessions would keep running. He was separately building cloud sandboxes that do not shut down. In the interview the system is described as a runner, which executes agents on a machine, plus a server with a uniform API, so a hosted, collaborative agent can be set up and customized. The excerpt covers their rationale, not independent results.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
videoThe Inference Frontier: from 100 to 10,000 tokens per second — Sean Lie, Cerebras CTOLatent Space
videoThe Future of AI Infra: from Kubernetes to Agent Sandboxes — Akshat Bubna, Modal CTOLatent Space
videoAI Agents Need Computers: 74% MoM Growth, 850K/Day Runs, & New Agent Cloud — Ivan Burazin, DaytonaLatent Space
Checking sign-in…
Loading comments…



