Loops, Skills, Artifacts, Workflows: How I Turned AI Buzzwords Into a Personal Assistant That Actually Improved My Life

Every AI conference throws the same four words at you: loops, skills, artifacts, workflows. Then it leaves them on the slide. Let's skip the hype and cash them in. I'll take them apart one by one — and I won't stop at four; I'll unpack a whole stack of AI buzzwords on real, practical tasks, not dry theory.

One worked example runs throughout: a personal AI assistant I've spent the past year building. It isn't a chatbot. It's a task queue with a roster of specialised agents that claim work, do it, and report back. Through that single build the jargon turns into concrete machinery. Loops re-run my recurring chores. Skills are just markdown files that teach an agent how I book my car or file my taxes. Artifacts are what one agent leaves behind so the next doesn't start blind. Workflows chain them together.

I haven't done my own life admin by hand in months. Car service, water meters, coffee re-orders, vinyl hunts, buying euros, power-outage graphs — the agents handle it. They even maintain themselves: every night they scan their own error logs, catch performance regressions, and rewrite their own skills. Some experiments worked brilliantly; others failed spectacularly. I'll show you both — including how this very talk description got brainstormed and submitted by the assistant it's about.