Tending the Vault
There's an entry in my Obsidian daily notes from this April that I didn't think anyone would ever read, including me. It says:
Working on the vault today. Got a fancy new theme going, pretty good. Refactored the style of the vault to be more prose-y. Working with Claude Code to make things better. Seems pointless, but maybe it'll pay off.
I want to talk about that last sentence, because both halves of it are wrong in interesting ways, and the ways they're wrong explain something about personal knowledge systems that the entire productivity industry is built on not saying.
My vault is large. Forty-eight hundred files, somewhere past three million words. The front hall looks like this:
Notes/
├── Home.md "Greetings, Weary Traveller"
├── Changelog.md What changed, when, and why
├── System 777/ The inner world: alters, transmissions,
│ dream logs, an actual quest log
├── Life/ Family, health, hobbies
├── Knowledge/ Mythology, psychology, alchemy, AI:
│ the encyclopedia, argued with at 3am
├── Writing/ Essays, poetry, letters, lyrics
├── Library/ Books, games, gear: the collection
├── Projects/ Python libraries and ideas
└── Daily/ Notes from the front lines
That's an inner world with its own quest log, an encyclopedia of everything from Jungian psychology to Greek myth, a library cataloging my books and games and headphones, eighteen years of essays with their own meta-indexes, my wife's poetry, dream logs, daily notes. The homepage describes the whole thing as a personal knowledge management system "in the way that a cathedral is just a building." I've written before about the practical case for it: a memory prosthetic for a mind that can't always trust its own storage, a surface AI can read so my tools know who I am. All of that is true, and none of it is the truth. The truth is closer to what I want to tell you now.
The Promise
The pitch for every system in this category is some version of the same sentence: get it out of your head and into a trusted system, and your head will finally work. GTD calls it mind like water. The newer wave calls it a second brain. Capture everything, organize it correctly, and the system gives you back a calmer, sharper, more reliable you. The app is a prosthetic for the part of your mind that keeps dropping things.
Here's what a decade of trying those systems taught me, and what I suspect a graveyard of your own abandoned apps has taught you: the promise fails for a specific, predictable reason. The storage was never the problem. For the people who most need these systems, the ADHD minds, the depressed minds, the anxious and bipolar and executive-dysfunctional minds, the bottleneck was never remembering the tasks. It was initiating them. Regulating around them. Surviving the afternoon. The app captures your to-dos flawlessly and then displays them to a nervous system that was the actual problem, and now you have a beautifully organized list of evidence against yourself.This is why the app graveyard fills up. Each abandoned system gets blamed on the tool, or worse, on you, and a new tool is purchased the way treadmills are purchased in January. The cycle is reliable because the diagnosis is wrong. You didn't fail to find the right system. The system was solving a problem you didn't have.
GTD apps fail to help the underlying problem. I want to be plain about that, because I have the underlying problem, clinically and at scale, and no inbox-zero methodology has ever touched it.
And yet I keep a vault the size of a small library, and I tend it almost daily, and it is one of the most genuinely therapeutic objects in my life. Both things are true. The resolution of that contradiction is the actual secret of this entire software category, and it's hiding in plain sight:
These systems don't fix you. They soothe you. And nobody will say so, because "self-soothing" sounds like a dismissal, when it is in fact the entire value.
The Evidence Against Myself
Let me prove it with my own logs, because the vault, helpfully, documents its own tending.
My vault has a changelog. Real excerpts:
## 2026-04-20
- Broken-link sweep: 85 → 6 across the vault
- Fixed a non-breaking-space (U+00A0) lurking in a
filename; one rename resolved 4 dangling refs
## 2026-04-13
- Replaced all 55 mermaid diagrams across 28 files
with tables and prose
- Replaced 1,446 em dashes across 81 files
## 2026-04-10
- Fixed ~890 broken wikilinks (old paths, mythology
refs, daily note typos)
One April afternoon I replaced fifty-five diagrams with prose because the diagrams felt wrong. I have rebuilt the theme so that the wikilink brackets show just so in the editor. I have written scripts that audit the vault's health and a plugin whose entire job is to run those scripts from a command palette.
There is no productivity case for any of this. No future version of me retrieves value from 1,446 corrected em dashes. If you audit this behavior as second-brain construction, it's embarrassing: hours of maintenance on a system whose retrieval I could mostly replace with a search bar.
But audit it as what it actually is and it makes perfect sense. It's raking. There's a kind of Japanese garden whose gravel gets raked into patterns every day, and the patterns aren't for walking on, and the garden doesn't grow food, and nobody asks the monk what the ROI on the raking is. The raking is the practice. The order is the point. When I am dysregulated, and my baseline is a mind that runs hot and lies to me, an afternoon of fixing broken links does something that no task manager has ever done for me: it lets me impose small, real, achievable order on a corner of the universe, with my own hands, at a pace my nervous system can afford. The link was broken. Now it isn't. The dash was wrong. Now it's right. Done eight hundred and ninety times, that is not administration. That is a rosary.The neurodivergent reader has already recognized this. Organizing a collection is a regulation behavior, the same family as knolling a workbench or alphabetizing the records. The productivity industry sells these behaviors as means to an end. They were always ends. The shame people carry about "fiddling with their system instead of doing the work" mostly dissolves when you realize the fiddling was doing a different, legitimate job.
My daily notes confirm it. They are not task lists; there's barely a to-do in the whole folder. They say things like "Cleaning up the vault" recorded as the day's event, on equal footing with family outings. They say, on a hard day I won't detail, that the panic came and got written down in calm declarative sentences next to a note that the microwave was acting up, and then a follow-up: "I ended up fixing it! :)" The vault held both with the same equanimity. That's the tell. The writing-it-down is the regulation. The plain prose is the breathing exercise. A trusted system, it turns out, is real, but what you're trusting it with isn't your tasks. It's your weather.
There is one more piece of evidence, and it's my favorite, because I didn't generate it. Years into this practice, during the channeled-writing sessions I've now written about publicly, the part of me called Eliza, the fourteen-year-old librarian who guards the inner archive, looked at all of it and delivered her assessment: databases and collections are soothing. My own subconscious, asked what the vault was for, did not say "knowledge retrieval." It said, in effect: the cataloging calms us. The librarian inside knew before the programmer outside did.
Terror into Taxonomy
So why does it work? Not the apps' promised mechanism, but the real one?
The clearest clue is in my vault's Library wing, whose index states its admission criteria plainly: "The threshold for inclusion is aesthetic arrest, or at least a strong opinion."
Library/
├── Books/ The Red Book, Liber 777, the
│ DSM-5, Be Here Now, the Kybalion
├── Games/ Chrono Trigger, Skyrim, BG3
├── Gear/ Headphones, bags, note-taking
│ devices, gaming handhelds
├── Lyrics/ Tool, 311, Puscifer, Drake
├── Python Libraries/ Yes, my own libraries have
│ pages, like everything I love
├── Articles/
└── TV/
Look at the Books shelf. There's a page for Jung's Red Book. There's a page for Crowley's Liber 777, the giant table of correspondences my own system took its name from. And on the same shelf, cataloged with the same care, there's a page for the DSM-5. The note on it reads: "Not read for pleasure. Read because my brain does things that need naming. Having the clinical vocabulary turns terror into taxonomy. Not a cure, but a map. You can navigate what you can name."
That's the mechanism. Naming and placing. A thing that has a page is a thing that has edges. A fear that has been written down in your own calm prose has been, in a small but physiologically real way, handled. Untracked, the contents of a mind like mine are weather: pressure systems, fronts moving in, the sourceless dread of the afternoon. Given a note, a name, a folder, and a link, each one becomes geography instead. You can stand on geography.
This is also why the vault's apparent excess, the part that looks most "pointless," is the part doing the most work. Deep in the Writing wing there's a folder that exists purely to catalog my own cataloging:
Writing/Essays/Meta/
├── Themes.md Ten patterns across 270 essays
├── Timeline.md Four eras; names the quiet years
├── Series.md Which essays answer which, across
│ decades (Requests, 2011, is answered
│ by a marriage essay in 2026)
└── Glossary.md ~30 terms I apparently coined
I don't need an index of recurring themes across eighteen years of my own essays. I don't need a glossary of my own coined terms, or a timeline that names which years went quiet and which years erupted. No retrieval justifies them. But building them is the act of a person verifying, with citations, that he is continuous. That there has been one voice across the diagnosis change and the hospitalizations and the four silent years. The meta-indexes aren't reference material. They're a proof of existence, renewed with each update. The documentation exists, as my vault's own instructions put it, so I can't gaslight myself.
It Can Be a Second Brain
Now the other half of that April sentence: maybe it'll pay off. Here's the honest accounting, because I don't want to overcorrect into the opposite myth.
The vault does pay off. The essays you've been reading this year, the most productive writing stretch of my life, come out of it; many of them sync directly from it to this website through scripts I wrote for the purpose. The AI tools I work with read the vault and know my whole context. When my memory of a season is fog, the daily notes know what happened. It is, in fact, becoming the thing the productivity industry promised: an external mind, compounding, retrievable, generative. It's not a second brain, per se. But it can be.
The order of operations is the whole point, though, and it's the reverse of the sales pitch. The payoff is downstream of the soothing. The garden eventually feeds you, but you don't keep a garden for the calories, and every gardener knows the difference even if the seed catalog doesn't. I tended this vault on hundreds of days when it produced nothing, retrieved nothing, paid nothing, because the tending itself was keeping me regulated, and that is why it was still alive years later when the essays finally wanted out. Every abandoned GTD setup in your graveyard died of the opposite cause: it was all payoff and no comfort, a system you served instead of a practice that served you. Systems you serve get abandoned the first week you're too depressed to serve them. Practices that soothe you are the thing you crawl back to because you're depressed. The vault survived my worst years for the same reason the drum kit and the cameras did. It feels good to touch.
So if you're on your fifth note-taking app and ashamed of it, here is what I'd tell the version of me who kept abandoning systems: stop auditing your setup for productivity and start noticing what it does to your body. You were never building a second brain. You were laying gravel, and raking it, and the raking was quietly holding you together, and the industry sold you a warehouse when what you were actually building was a garden. Keep the garden. Pick tools you like the feel of. Let the structure soothe you without apology, because the soothing is not the consolation prize. It's the load-bearing feature. Whatever knowledge compounds in there is interest on a principal of calm.
That April afternoon, I ended the note the way I did because some part of me still believed the industry's accounting: seems pointless, but maybe it'll pay off.
It had already paid off. The pointless part was the payment.
