I Gave an AI Agent the Keys to My Life (Here's What Happened) — Radek Sienkiewicz (@velvetshark-com)

An honest look at what happens when a personal AI agent is allowed to operate around the clock. Over months, one permission at a time, it went from reading files to handling email, backing up its own memory at 2am, monitoring its own health, and drafting real business replies. This talk covers the permission creep, the overnight cron ecosystem, self-monitoring and recovery, trust boundaries, and the surprising value of giving an agent a personality that disagrees with its owner. Speaker info: - https://x.com/velvet_shark - https://www.linkedin.com/in/radeksienkiewicz/ - https://github.com/velvetshark Timestamps 0:15 Radek's path to OpenClaw 2:17 The philosophy of incremental growth and system updates 4:51 Integrating the Obsidian knowledge base 8:59 Ambient operations and overnight automation 11:02 Core job types for the AI agent (Ambient Operations, Attention Filtering, Execution) 13:03 Deep dive into specific Discord integration channels 14:54 System architecture: LLMs, scripts, and memory management 16:28 Challenges: Bad memory, brittle automations, and noisy nodes 17:19 Conclusion: Optimizing for the future self