Descriptions:
Peter Yang walks through a complete setup guide for OpenClaw (also called “Claude” in the video), a personal AI assistant built on Claude Code that operates through messaging apps like Telegram and can execute real computer tasks autonomously. The tutorial covers both capability demonstrations and the practical installation steps needed to get a 24/7 AI assistant running safely.
Live demos include updating a personal website via a Telegram text message (the change propagates to GitHub and auto-deploys via Vercel in seconds), sending a personalized email through Gmail, summarizing a Twitter timeline without doom-scrolling, posting and deleting a live tweet, and reading goal documents from Google Drive to inform personalized advice. Yang also shows how OpenClaw’s memory layer reads planning documents so the assistant understands priorities and context without repeated explanations.
For setup, Yang recommends three hardware options — an existing old MacBook, a Hetzner VPS ($5–10/month for 24/7 uptime), or a dedicated Mac Mini (~$500). A key safety section advises running OpenClaw on a machine with its own Apple ID and Google credentials, with access scoped only to specific calendars and Drive files rather than full account access. The video closes with community use cases: health tracking via Whoop integration, automated flight check-ins, and smart home control. It’s a practical starting point for anyone wanting to build a persistent personal AI agent using Claude Code.
📺 Source: Peter Yang · Published January 28, 2026
🏷️ Format: Tutorial Demo
