I Built an AI Journalist That Embedded With My Life

I’ve been thinking a lot about what AI is genuinely good at beyond writing code and answering questions. One pattern keeps surfacing: AI is excellent at the slow, low-priority work humans never get around to. The stuff that would be useful but never important enough to actually do.

Journaling sits squarely in that bucket for me. Twenty-six years in network engineering and I have approximately zero personal journal entries to show for it. So a few weeks ago I tried something different. Instead of trying to journal more consistently, I’d hire an AI to do it for me. Not transcribe what I say into a notebook — actually observe my life and write about it the way a beat writer covers a player.

I called the project Cronkite, after the news broadcaster. The idea was simple: give an AI the brief of an embedded reporter, point it at my life, and let it produce a daily piece of writing in a notebook-style journal I could read back later.

What “Embedded” Actually Means

For an AI to embed with your life, it needs triggers. Things that make it pipe up and start writing. Cronkite watches for three.

Sports games for teams I follow — when there’s a meaningful Phillies, Eagles, or Aston Villa result, it knows I probably care about how that went.

Conversation follow-ups — if I mentioned something the day before that has a natural arc to it (the home router rebuild, an upcoming trip, a project that was stuck), it checks back in on it.

Random buddy thoughts — sometimes it just sends a text. Like a friend who hasn’t seen you in a few days asking what’s up.

The pacing matters more than I expected. During the calibration window I capped it at one trigger per day. Any more and it stops feeling like a reporter and starts feeling like another notification source begging for attention. Less than that and the journal goes empty.

What Shows Up On The Page

The actual entries surprised me. They’re written from a slightly external perspective — observing what I did and what happened — but they’re not lifeless transcripts of “today Brad worked on his network.” They have a voice. They notice patterns across days. They remember what was unresolved from the last entry and bring it up in the next one.

A standard entry might note that I spent most of the day on a particular project, that the home network has been acting up since the router swap, that an old friend texted, and that the Eagles game was on in the background. That sequence of facts, written the way a sportswriter would write them, becomes a real entry I can read back later and remember what last week actually felt like.

This is the part that gets me. I have months and months of activity logs across my projects — commits, agent sessions, nightly snapshots — but they don’t add up to memory. They’re records. Cronkite is the layer that turns records into a story.

Shipping The Reader

The journaling part worked from the day I switched it on, but the entries lived in plain markdown files on a server I had to dig into to read. Functional, but nobody reads a journal that way.

So this week I shipped the front end: a calendar grid that shows which days have entries, a day view that renders each entry into a notebook-paper aesthetic with proper serif type, and the whole thing pinned to my private network so nobody else can see it. It runs as an always-on service alongside everything else, so I can pull it up from my phone or laptop and just scroll through what the reporter wrote about my month.

The user-experience bar was low — I just needed something better than opening text files in an editor. But what I got is a journal I’ll actually open. And that’s the whole point.

Why It Works

The thing that makes this useful instead of cute is the persistence layer underneath. Cronkite isn’t a chatbot you open. It’s part of a 24/7 system that already knows what’s happening across my projects, what I’m working on, what’s been said and not said. It’s not generating journal entries from a blank page. It’s writing about a life it’s already in.

If you’re thinking about whether AI assistants could do this kind of quiet, persistent, low-priority work for your own life or business — observation tasks that humans never get around to — I’d be happy to talk through it. Head over to /ai-assistant to see what I’ve been building, or hit /contact and tell me what you’re trying to figure out.