Summaries are for agents.
Humans think in stories.
story-rex takes the recordings of a life (meetings, phone calls, a pocket recorder left running at a card show) and weaves them into literary stories in an Obsidian vault. It links and tags them, and checks every quote against the recording before a story counts as done. This page tells how it came to exist, following the same arc the vault tells about itself.
- The group visited a vineyard and tasted wine.
- The host discussed vineyard operations and local history.
- Action items: none.
This is the summary that started story-rex. Two weeks later Chad told his brother about this recording: several conversations going on at once in between different people… a summary can't do that.
Stump Town. A Saturday tasting at Porter-Bass, a small family vineyard, with strangers who become company for an hour. Shape: ensemble.
The host runs the place with her son and introduces herself mid-pour, Oh, my name's Sue,
after Chad admits he should have caught it earlier. He wants to know how a vineyard this size gets worked. So my son and I can cover all the tractor work and the running, the place, irrigation and spraying and everything. But we have a crew that we get in to do hand work in the vineyard.
Someone at the table asks about the history, and Sue gives the afternoon its name:
So this was all redwood forest that they logged in 1850.
That's why they call it Stump Town.
Five voices, one microphone, and no names in the labels. The story only knows who Sue and Chad are because each of them said so.
Stories, not summaries
The pitch, made to his brother on a Saturday evening.
It started with a habit of pressing record. Granola took the meetings, Wispr Flow took the dictation, and the transcripts piled up. On a Sunday call on May 10, Chad told his brother what they were for:
I'm not going to call it a weapon. But ammo… it's transcripts and my thoughts.
He wanted a data set that is Chad… then I can extend myself.
By May 16 he knew what to do with it. Granola's summaries were garbage on messy audio, and the tasting at the top of this page was the proof: two hours, overlapping conversations, strangers joining in. A summary can't do that.
What did work was asking an agent to write it up as a story and link it into an LLM wiki. I think it's the most powerful thing I've ever seen AI do.
The working name Claude gave it was Storify. That evening he called Sandy and said it out loud:
I don't want to focus on summaries, I want to focus on stories.
Sandy put its odds as a business at maybe two percent and gave it full marks as something worth doing. Chad answered with advice he'd taken from founders he admires: build something that you love, and the rest will fall into place.
The vision, as first stated: not a product, but the interface that lets everyone keep their own version of one story… when everything is recorded, that's the future I'm trying to design for.
It comes back at the end of the page.
Inside day_watcher
Every session gets a folder. There were 137 of them by Sept 20. Pick a day to open one.
What it can become
Each idea below already exists as a seed somewhere in the vault. Filter by how far off it is.
Where the arc points: the May 16 vision was everyone keeping their own version of one story. Each item above moves a vault built around one person a step closer to that.