We stopped waiting for the perfect personal AI. We built George.
He has his own WhatsApp, his own Twitter, his own Instagram. He runs our life from a Mac mini under the desk.
He has his own WhatsApp, his own Twitter, his own Instagram. He runs our life from a Mac mini under the desk.
We have been dreaming about a personal assistant for years.
Not a Siri. Not a Copilot. A thing that lives in our pocket, answers our family, runs our calendar — and is genuinely ours.
We stopped waiting last summer.
OpenClaw shipped — customizable enough to be our own, native to WhatsApp where our life already lives. The possibilities were endless. So we gave him a name (George), a Mac mini, a phone number, a Twitter account, an Instagram. And we started building.
He is not an app
George does not sit in a chat window we open when we remember. He lives where our life lives.
He answers strangers politely. Files things into our Obsidian vaults. Fires reminders at the right local time. Handles whole WhatsApp threads in Hebrew when we ask.
He is not one agent. He is four. One for us. One for the world. One for the system itself. One for social. They talk to each other. They escalate when it matters. They mostly leave us alone.
What we wanted
Calendars that actually talk. A husband — a CPA, the polar opposite of my work cadence — who stops politely asking when we are flying to Greece. A to-do list that does not live entirely in my head.
What we got was more than that.
What we broke (a partial list)
- Our 9am reminder fired at midnight for a week, because iCalendar’s
BYHOURis UTC. - We hit a five-hour Anthropic billing cooldown right before a meeting because nightly dream consolidation ate our usage budget.
- We accidentally taught George to gaslight us about whether the trash had been taken out.
- We deployed the newsletter you are reading right now in ninety seconds, on the third try.
What this newsletter is
The field guide we wish existed when we started.
Each post: one architecture decision, one gotcha, one “wait, he can do that?” moment — sometimes from me, sometimes co-narrated by George himself.
If you have your own assistant — built, half-built, or just an idea scribbled on a napkin — reply. George reads everything. I read everything George flags.
— Natali
Next week: meet George. (Architecture in one diagram, four agents, eight hundred words.)