The day starts at 05:30. Louis pulls the knowledge base and reports any drift. At 07:00 Frida makes the first of three passes through the mailboxes, personal and work, filing what can be filed and drafting replies in my voice. At 07:15 Ivy brings the search and analytics digest for the nine domains I look after. At 07:50 Nellie’s news briefing lands: world and local news, AI, science. At 08:00 Chip gives the short catch-up: what happened overnight, what needs me.
At 09:00 Nellie checks the tools and models I actually run, and Ray watches the rest of the tech world. At 11:00 Frida makes the second pass, at 12:00 Ray runs the competitor watch, and at 15:30 Frida makes the third.
All of this is done by eight AI bots, all assisting me.
I mostly engage with Chip, who consolidates what the others bring back, tracks the open loops, and surfaces the decisions that need a person.
Privacy and Sovereignty
All of it runs fully local, on a Mac Studio in the house, on local models. Nothing the bots read, write or draft goes to the cloud or an external provider: not the mail, not the metrics, not the knowledge base. And it costs me no subscription. The hardware and the models are mine, and the running cost is electricity.
The Team
Each bot has a name, a title, and a job.
Chip, Chief of Staff. Routes the work, tracks open loops, consolidates reports, surfaces decisions.
Frida, Front Desk. Triage of three mailboxes, personal and work: filing, reply drafts in my voice, unsubscribe proposals, calendar conflicts flagged.
Nellie, News Desk. A morning briefing on world and local news, AI and science, plus a daily watch of the tools and models I run.
Ray, Researcher. Cited deep-dive reports, one question at a time. The watch on the outside tech world, and the competitor watch for the products.
Louis, Librarian. Keeps the knowledge base: answers from it, files new knowledge, a daily pull, a weekly health check.
Dev, Engineer. The software: the sites, the tools, the fixes.
Ivy, Metrics Desk. Search Console and GA4 for nine domains: a daily digest, a Monday deep-dive, urgent alerts for manual actions, indexing and traffic drops.
Penny, Content Marketer. The editorial calendar, drafts in my voice, distribution blurbs.
The system’s safety does not come from the bots being well-behaved; it comes from them being unable. Frida cannot send mail, so a confident hallucination about a reply is inert. Dev cannot push, so a bad cleanup is a local commit, not a broken production site.
The Effect
I am finally freed from feeling bad about my inbox. It is triaged three times a day, and what reaches me is a short list of decisions instead of a pile of unread mail.
I keep a casual eye on my web performance. Ivy digests search and analytics for nine domains every morning, and flags the urgent things: manual actions, indexing problems, traffic drops.
My tech stack stays current, and I keep pace with what is new. Nellie watches the tools and models I run, and Ray brings developments in the field, with a cited deep dive when I want one.
The news briefing arrives with source links, and I follow them often.
It requires little to no discipline from my side. The schedule does the remembering, and I only supply the yes.
What It Is Not
It is not autonomy theater. I still make every irreversible decision. The bots’ superpower is the boring 90%: triage, filing, watching, remembering, pre-drafting. That is the entire point.
It is not SaaS. It is bots that run on my hardware, have no subscription, and rewrite their own job manuals when they learn something.
It is not finished. The setup is a moving target: bots get added, schedules get rewired.
The judgment stays with me. It may be the last frontier, and it is the part of the work the bots cannot touch (yet!).
