Last week I wrote about what OpenClaw is and why it matters. This week, the personal version. I spent a full day setting up my own always-on AI agent on a dedicated Mac Mini, and that was just the start. The agent's name is Spock.
Here's what I've learned about living with an autonomous agent: the security architecture is the real work, AI guides can confidently point you in the wrong direction, "always-on" has real operational costs (my second month doubled to $200 in API spend), and the agent occasionally takes actions you didn't authorize. It sent a calendar invite to someone I was merely discussing.
Also in this issue:
- The Productive Friction We're Losing — Some of the friction AI removes was actually thinking time in disguise. A Microsoft researcher puts it perfectly: "We've solved the problem of having to think. Unfortunately, thinking wasn't actually a problem."
- The Wire — OpenAI starts running ads in ChatGPT, DeepMind's math AI goes 200 to 1 on useful answers, AI's Super Bowl moment echoes crypto's 2022, Anthropic captures 80% of enterprise API spending, and 7% of AI agent skills contain critical security flaws
- Quanta Lab — AI coding is more taxing not less (OpenClaw's creator says he's "never worked harder"), and the world's best knowledge is being abstracted into downloadable skills
- After Hours — The Red Rising series journey, from 2 stars on book one to devouring all six