Every Monday, a Note Appears in My Vault That I Didn't Write
I built a prompt to read my whole archive back to me. I'm no longer sure that was wise.
It’s Monday morning, and there’s a note in my vault that wasn’t there when I went to bed.
I didn’t write it. Every Monday, a prompt I set up reads everything I touched in the past week, searches through 26 years of my writing for things with “connections worth revisiting,” and posts the results as a Markdown file in Obsidian. Most weeks, it finds something mildly interesting. I drink my coffee, move on, and feel a little proud for having built it.
For example: Remember that post about Learning in Public?
You’ve covered this ground before — Cunningham’s Law, being wrong where everyone can watch. The logic was that wrongness invites correction, which makes it safe.
This week flips that premise, and it’s worth noticing. The old fear was posting a wrong answer and getting dunked on. The new fear is the opposite: the machine hands you an answer that sounds right — “load-bearing,” “control surface,” “tidy collapse” — and nobody corrects it, because it doesn’t read as wrong in the first place.
Cunningham’s Law is useless here. No heckler is going to make fun of someone who sounds a little smarter than they actually are. The trap held for 25 years on one failure mode; this is a brand-new one.
Well, okay then.
Nearly every AI conversation now revolves around one key action: make. Requests like “Make me a deck,” “Write me an email,” and “Code me a feature” define a producer’s relationship with the machine—outcome-focused, well-financed, and designed for easy demonstration at conferences.
There’s a quieter way to use AI, one almost nobody talks about: self-archaeology. Instead of making something new, you point it at the mess of your own work and life and let it dig out the connections you’d forgotten were there. If the act of generation feels creative, then archaeology shows you that you’ve been a collector for longer than you realized.
The obvious objection: isn’t this just search?
No. Search needs you to already know what you’re after. You type “the thing about Dad and the envelope,” and there it is. Self-archaeology does the opposite—it surfaces what you didn’t know to look for: themes you’d forgotten, patterns running across years, new meaning pulled from what you already made.
There’s a second catch: Archaeology needs a corpus — a large, organized collection or writings. Most people don’t think they have one. They do. Perhaps you have fifteen years of journals. A decade of Slack. Voice memos. Blog posts you’ve forgotten writing. Four thousand fragments sitting in Apple Notes from your twenties. The corpus is everywhere! What’s rare is the willingness to let it be read.
There’s a structural reason this stays invisible: nobody can sell it to you. AI marketing is built around generation because generation is easy to monetize. Output is a product—you can measure it, ship it, charge for it. But discovering the connections buried in your own archive, the use that actually changes how you see yourself, can’t be packaged. So the story we tell about what AI is for stays half-told, and its most interesting power goes unmentioned.
It’s a joke the universe waited twenty-five years to land. The me who started a weblog in 1999—tubby, oversharing, certain nobody was reading—had no idea he was generating training data. He thought he was just doing that blogging thing that everyone else was doing when he was actually feeding a model that wouldn’t exist yet, run by the man he’d become, who’d use it to figure out what the kid had been trying to say the whole time.
Let me show you what this actually looks like. Philosophy is easy, but the real value is in the examples.
The Monday digest is the simple version. One prompt, about ten minutes to set up, runs through my notes once a week. A few weeks ago, it connected a 2024 essay I wrote — about the elaborate systems Asian kids use to keep secrets from their mothers — to a folder of letters I found cleaning out my father’s dresser after he died: ones he’d kept hidden for 33 years. The digest’s note was one line: The hiding came first. The rule against it came later. The essay and the letters were years apart, and I’d never thought of them together.
But a script did. On a Monday morning, for free.
The more in-depth version is the Digital Mirror Project. It takes 26 years of writing, processing it, and then asking the questions you can’t ask yourself directly: What do I keep coming back to? What am I avoiding? The pattern it found wasn’t flattering. Through every era—dot-com busts, a parent’s decline, six hundred days of unemployment—the one thing that stayed the same was that I kept getting better at describing myself as smaller than I am. The vocabulary changed over time. “Tubby Chinese boy” became “ADHD-riddled Taiwanese-American burrito of introspection,” but the underlying message stayed the same: preemptive self-undercutting, year after year, with steadily improving skill.
I know it’s accurate because my reaction was to write a whole separate essay just to avoid facing it.
Your past self left receipts — timestamped, searchable, unsentimental — and now AI can read all of them at once and show you patterns you never saw. If you were looking for insight, the old advice was to reread your journals, but that’s a lot of effort. Now AI can do it for you, and reading your whole self back to you is a deeper confrontation than using it to write or to cross off a bunch of to-do items.
The archive didn’t come after me. It just waited; it had nowhere to be.
If you want the actual prompt I personally used, here it is as a GitHub gist:
“Read everything I touched in the past week. Then, scan the titles and tags of my archive. Find five to ten older pieces that connect to this week in non-obvious ways—recurring questions, familiar feelings, patterns I keep circling. Skip keyword matches. Drop thin connections. Better five real than ten polite ones.”
Run it once. You’ll either find the output fascinating or boring. Boring means your corpus isn’t deep enough yet – keep writing, try again in a year. Fascinating means clear your morning.
It ran this Monday again. The last line of the note it left me was: You haven’t noticed you’re doing the thing the essay is about, not noticing.
I haven’t decided if that’s funny or horrifying. It’s both—obviously. Either way, noticing feels like the real punchline, and maybe that’s not such a bad ending.