Personal AI is Already Here (And You’re Probably Not Ready)
Summary
Robert Glaser documents a three-week experiment running OpenClaw on a Raspberry Pi to act as a personal AI assistant, illustrating that frontier models are now constrained by integration, data access, and security. The piece argues for a skills-based architecture, data ownership, and guardrails as essential for practical use, and it outlines concrete home-use workflows and the security risks involved. It also offers guidance for developers building agent ecosystems.