JustJot.ai
← Articles
ai-notetaking2026-06-17

"Why You Should Delete Most of Your Notes"

"Everyone tells you to capture more and never delete — storage is cheap, so hoard. But an archive that only grows is a liability, not an asset. The skill that makes a note system usable isn't capture. It's pruning."

the contrarian

Why You Should Delete Most of Your Notes

The advice is everywhere and it sounds responsible: never delete a note. Storage is cheap, search is good, and you can't know today which scrap you'll need in two years — so keep everything. Capture widely, archive forever, and trust that a bigger pile is a richer you.

I want to argue the opposite. An archive that only ever grows is not an asset. It's a liability you've been told to feel good about. The note you can't find is worth nothing, and every note you keep makes the ones around it slightly harder to find. The skill nobody sells you — because it doesn't involve a shiny new app — is pruning: deciding, on purpose, what to throw away so that what remains can actually be used.

This is a deep-dive on how to do that without losing the one note you'll regret. By the end you'll have a deletion policy, a weekly pruning ritual, and a clear rule for the rare note that really is worth keeping forever.

TL;DR - A note's value is its retrievability × usefulness, not its existence. Unfindable notes round to zero. - Every note you keep imposes a small tax on every search, every review, and every link you draw. Hoarding has a cost; it's just invisible. - Steelman accepted: capture should stay cheap and wide. Pruning happens after capture, not instead of it. - Use the Keep / Compress / Cut triage and a 10-minute weekly pruning pass. Default to cut. - Keep forever only what's durable, hard to re-derive, and yours — decisions, distilled principles, original work. Delete the rest without guilt.

The consensus view — and what's right about it

Let me steelman the hoarders before I dismantle them, because their core instinct is sound.

The "keep everything" camp is reacting to a real pain: losing a note you needed. Anyone who has deleted a file and wanted it back a month later learns the lesson hard, and over-correcting toward keeping everything feels safe. They're also right that capture should be frictionless — if you hesitate at the moment of capture, you lose ideas, and that loss is silent and total. And they're right that storage is, literally, cheap: a million words of notes costs you nothing on disk.

So concede the strong version of their case:

All true. If "keep everything" meant only "don't filter at the moment of capture," I'd have no argument. My argument is with the second, sneakier claim hiding inside it: that because storage is cheap, keeping is free. It isn't.

The hidden tax on every note you keep

Here is the move the consensus skips. Storage is cheap; attention is not. A note doesn't sit inertly on a disk — it sits in the path of every future search, every review, every time you scan a folder looking for something else. The cost of a note isn't paid once when you save it. It's paid in tiny increments, forever, by your future self.

What grows with your archiveWhat it costs you
Search results per queryMore near-duplicates to read past to find the real one
Items in any review passReviews get longer, so you do them less, so the system rots
Links/backlinks to maintainYour graph fills with weak connections that dilute the strong ones
Decision fatigue at retrieval"Which of these six similar notes did I mean?" — so you trust none

This is why the 10,000-note vault so often feels worse than the 200-note one, even though it "contains more." More is not richer when the signal is buried. A note system isn't a warehouse; it's a tool you reach into under time pressure, and every dud you keep is one more thing your hand brushes past before it finds the thing you wanted.

This is the same mechanism behind the [collector's fallacy](what-is-the-collectors-fallacy.md): saving feels like progress, so we over-save. Pruning is the missing other half — the part that turns a collection back into a usable tool.

Worked example. You search your notes for "pricing strategy." A pruned vault returns one distilled note: your three principles plus a link to the source. A hoarded vault returns nineteen results — four article clips you never read, two near-identical meeting notes, a screenshot with no context, your actual distilled note, and eleven things that merely contain the word "price." You found the same answer in both. In one of them you spent forty seconds and some willpower to do it. Multiply that by every search you'll ever run.

The retrievability test: what a note is actually worth

Replace "might I ever need this?" — which always answers yes — with a question that has teeth:

Value ≈ Retrievability × Usefulness-when-retrieved.

Both terms can be near zero. A note you'll never surface (no title you'd search, no link, buried in a daily log) has retrievability ≈ 0, so its value rounds to zero no matter how good it is. A note you'd surface easily but that just re-states something obvious has usefulness ≈ 0. "Keep everything" implicitly assumes both terms are always positive. They usually aren't.

So before you keep a note, ask the two questions the hoarder never does:

  1. Future-me test: Will I plausibly search for or stumble onto this again? If you can't name the situation that would surface it, it's already lost — deleting it just makes the loss honest.
  2. Re-derive test: If I needed this and didn't have it, how hard is it to get again? A fact one search away is cheap to lose. A hard-won decision, a distilled lesson, or your own original synthesis is not.

A note worth keeping clears both: you'd retrieve it, and you couldn't cheaply rebuild it. Most notes fail at least one. That's not a tragedy — it's the system working.

The Keep / Compress / Cut triage

Deletion doesn't have to be binary. Most "should I keep this?" decisions resolve into three moves:

The Keep / Compress / Cut framework - KEEP (as-is): Durable, hard to re-derive, and you'll retrieve it. Decisions, principles, original work, reference you actually reuse. Rare — aim for the minority. - COMPRESS: The insight is worth keeping but the bulk isn't. Distill the article clip to its one transferable idea and delete the clip. Merge five meeting notes into one decision log. This is where most "keep" instincts should actually land. - CUT: Failed the retrievability test, the re-derive test, or both. Duplicates, stale to-dos, context-free screenshots, "interesting" links you never opened. Delete without ceremony.

Compression is the release valve that makes aggressive cutting safe. You're rarely choosing between keep the whole thing and lose the idea forever — you're choosing to keep the idea and cut the packaging. (If you've read [progressive summarization](what-is-progressive-summarization.md), compression is its terminal step: the final distillation is the keeper; the raw highlight underneath it usually isn't.)

Worked example. You saved a 3,000-word article on sleep. The Keep instinct says "archive it." Compress instead: write one line — "Consistent wake time matters more than total hours (source: [author])" — and delete the clip. You kept 100% of what you'll ever use and 0% of the search-clutter.

A 10-minute weekly pruning ritual

Pruning fails when it's a someday project — "I'll clean up my vault when I have a free weekend" never arrives, and by then the pile is intimidating enough to guarantee it stays. The fix is the same one that saves note-taking: make it small, regular, and boring. Bolt it onto a review you already do (see [the weekly review](../productivity/the-weekly-review.md)).

Weekly prune — 10 minutes, default to cut - [ ] Open notes created in the last 7 days. For each, run Keep / Compress / Cut. - [ ] Delete the obvious duds first (context-free clips, dead to-dos, dupes) — fast wins build momentum. - [ ] Compress the "I want to keep this" pile: extract the one line, cut the rest. - [ ] Tag the genuine Keepers so they're findable (a title you'd actually search beats a clever folder). - [ ] Stop at 10 minutes. Pruning is a habit, not a purge.

The "default to cut" rule matters: when you're genuinely unsure, lean toward deletion. The asymmetry favors it — most notes you cut you'll never miss, and the rare one you do can usually be re-derived in a search. Keeping a doubtful note costs you a little forever; cutting it costs you nothing unless the unlikely happens. Bet with the odds.

What actually deserves to be kept forever

So pruning isn't anti-note. It's pro-signal. The point of cutting the 90% is to make the 10% legible. Here's the 10% — keep a note forever when it is all three:

TestKeep-forever noteDelete-it note
Durable (won't go stale)A principle, a finished decisionThis week's to-dos, a price quote
Hard to re-deriveYour own synthesis, a hard-won lessonA fact one search away
Yours (you made meaning)A distilled insight in your wordsAn unread clip in someone else's

Decisions and their reasoning. Principles you've distilled from many sources. Original work and the thinking behind it. The handful of references you genuinely reuse. That's the permanent collection — and it's small, which is exactly why it's valuable. A vault of 300 notes you trust beats one of 10,000 you don't.

Common mistakes

Summary + next step

The honest version of note-taking has two skills, not one. Capture gets all the attention because it's easy to sell and easy to do. Pruning is the half that makes the other half pay off — it's the difference between a tool you reach into with confidence and a warehouse you avoid. Keep capture cheap and wide; then, every week, run Keep / Compress / Cut and default to cut. A note's worth is its retrievability times its usefulness, and most notes fail that test. Letting them go isn't loss. It's how the keepers become findable.

Next step: pair this with [how to organize your notes](how-to-organize-your-notes.md) — once you've pruned to signal, a light structure keeps it that way. In JustJot.ai, lean on search and links over folders so the few notes worth keeping surface the moment you reach for them — and let the rest go.