Structure Beats Magic ← All writing

For knowledge workers

I Wired Twenty Data Sources Into My Daily Notes (And My AI Reads Them)

A daily note that you have to fill in by hand is a chore you'll abandon by February. So I stopped filling mine in. Twenty-odd data sources — mail, calendar, banking, location, photos, reading, health — write it for me. Here's the actual machinery, not the pitch.

By Jaco van der Laan · 2026-07-04
I Wired Twenty Data Sources Into My Daily Notes (And My AI Reads Them)
Twenty-odd data sources — mail, calendar, banking, location, photos, reading, health — flow into structured stores, get linked and enriched, and assemble themselves into the day's note. I stopped filling it in; my data does it. Private by design: you own your data.

The daily note nobody keeps

Everyone who's tried journaling or a "daily note" habit knows how it ends. Week one, diligent. Week three, sporadic. Week six, a graveyard of empty dated files. The problem isn't discipline — it's that you're asking a human to be a data-entry clerk for their own life, every single day, forever. That never survives contact with a busy week.

So I inverted it. My daily note isn't something I write. It's something my system assembles — from data I'm already generating just by living. Where I was, what I spent, what I read, what I photographed, who I mailed. I show up to a note that's already half-written from facts, and I add only the part a machine can't: what it meant.

This is the enrichment layer, and it's the least glamorous, most valuable machinery I own. Let me open the hood.

Show the machinery: what actually feeds the note

This isn't a concept. It's a running system, and here are real numbers from it as I write:

And more around the edges — WhatsApp, CRM, trips, LinkedIn, Notion, published articles. All told, on the order of twenty sources, each one a small importer that lands data into a local DuckDB database, and roughly twenty connectors that let my AI query them live.

None of it was a grand project. Each source was a single afternoon: pick a thing my life already produces, write the small importer that captures it, move on. The system is the accumulation of those afternoons — twenty-odd of them by now — which is exactly why it compounds instead of sitting there as a plan.

How a fact becomes a line in a note

The pipeline is duller than it sounds, and that's the point — dull means reliable. Roughly:

1. Sources land in databases. Each source has an importer that normalizes its export into a local DuckDB table. Mail, calendar, transactions, location, photos, health — one database per source, kept separate. 2. An enricher assembles the day. A script reads a day's date, queries the relevant databases, and writes structured sections into that day's note — where I was, what I spent, what I measured, what came into the inbox that mattered. 3. It cascades up. Days roll into weeks, weeks into months, quarters, the year — so the summaries at every zoom level stay current without me touching them.

The note isn't a container I fill. It's a view, generated from my own data, that I then annotate. The facts are the machine's job. The meaning is mine.

From data to preferences — the part that surprised me

Filing facts is useful. But the machinery does something I didn't expect when I started: it derives preferences — a model of me — from the exhaust of ordinary life.

A pattern-extractor reads back over the enriched notes and surfaces the recurring things: the restaurants I return to, the places, the brands, the people, the kinds of book. Facebook's export gives follows and likes; reading highlights show what I actually engage with; transactions show where my money truly goes versus where I say it does. None of that is me sitting down to write "here are my preferences." It's derived — inferred from what I already did.

That's the difference between a profile you fill in (aspirational, stale in a month) and a profile that's observed (true, and always current). The system knows my tastes because it watched, not because I declared.

The honest state of it (because Proof means the warts too)

I'd rather show you the real system than a shined-up demo, so here's what's not perfect:

I'm telling you this because a system described as flawless is a system being sold, not shown. The value isn't that it's complete. It's that the architecture is right — sources into structured stores, structure into an enriched timeline, the timeline into a derived model of me — and any gap is just one more afternoon's importer away from closing.

Why this is the same lesson, again

Strip it down and it's the thesis of everything I build. A pile of raw exports — mailbox dumps, location JSON, transaction CSVs — is noise. The same data, routed through structure into a daily note that sits on a timeline, becomes a queryable memory of a life. Same bytes. Opposite value. The fork is structure, chosen on purpose.

And it's why the AI is useful to me in a way it isn't for most people. It's not that I have a cleverer model. It's that mine can see — twenty sources of real, structured data — and the seeing is what turns a generic assistant into one that knows where I was last Tuesday and what I've been reading all year.

Yours to take, and the part that isn't

The pattern here is free and I'll say it plainly: get your own data out (almost everything has an export — Google Takeout, your bank, your reading app), land each source in a structured store, and generate your daily view from it instead of typing it. Take it and run.

The kitchen — the part that's genuinely hard — is making twenty sources cohere: one database per source without it becoming a swamp, an enricher that assembles a day cleanly, a timeline that summarizes itself, a preference model that stays honest. That's not a script you paste. It's an architecture, tuned over years to one person's actual life. Designing that — for you, or for your organization's data — is the work worth doing together.

Because in the end the daily note that writes itself isn't magic. It's twenty boring importers and a spine to hang them on. The magic was never the model. It's the machinery — and the machinery is just your own data, given a deliberate shape.

Structure + Data + AI + Rules + Skills → Systems

← More writing Work with Jaco →