Open Source AI Agents on User-Owned Infra: K3s, MCP, & GPU Sharing in Practice
summary
An open-source personal cloud OS runs on single-node K3s personal cloud, serves models locally (qu 3.6 27B tested vs. Claude 4.5 Opus), hosts private knowledge bases, and enforces per-app sandboxes with observable, revocable connections; GPU sharing uses time-sharing GPU and memory reslicing to let competing AI workloads coexist; remote access runs over Tailscale + Cloudflare tunnel so data never leaves the device. Demo shows app-store installs, automated lifecycle with restarts, a 20-link daily Hacker News agent, and some demo flakiness but core flow works.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:03AI agents rise and challenges
- 3:45Security and privacy risks
- 7:20Local AI deployment necessity
- 8:44Open-source AI OS
- 12:19AI agent app store setup
- 15:17GPU sharing and resources
- 17:30Sandboxing and security
- 19:01CLI tools and control
- 20:07Live demo intro
- 21:40Deployment options and hardware
- 23:06Data sovereignty and cost savings
- 25:21Open-source vs frontier models
- 27:08Local AI tradeoffs
- 28:33Marketplace and agent demos
- 30:01Live agent workflows demo
- 33:45Conclusion and Q&A
related talks
07mOwn your Agentic AI
29mOpen Source Sovereign AI: Building Enterprise GenAI Platforms for Control & Comp...
46mAI Agents Are Becoming Operating Systems: Lessons From OpenClaw
15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?
01h 42mBuilding Responsible AI Agents in Decentralized Open-Source Environments
17mDevOps for Private AI: A Scalable Platform Using Kubeflow
20mOperating Frontier Intelligence at Scale
30mBuilding a Self-Service Platform for Stateful AI Agents