Note collections do not fail all at once. They decay quietly, through vague titles, dead links, and captures that never got finished. The fix is not constant gardening. It is a small set of repairs made while you are already handling the material, plus one short reset each month that you are allowed to stop.
- Diagnose the specific decay: titles, sources, backlog, or blurred attribution.
- Rename notes to state their claim the moment a vague title slows you down.
- Capture link, author, and date at the point of saving, never afterwards.
- Repair notes while they are open, and leave unused ones alone.
- Time-box the monthly reset and stop when the timer ends.
How a collection actually decays
Four failures cause nearly all of it. Titles stop describing content, so search returns nothing useful. Sources vanish or go behind a paywall. Half-finished captures accumulate until the inbox becomes unreadable. And your own paraphrase blurs into quoted material, so you can no longer tell whose words you are holding.
None of these is fixed by reorganising folders, which is why an afternoon of tidying leaves the collection no more useful than before. Each failure has its own small repair, and working out which one you are looking at is most of the job.
Title notes by meaning
A title should say what the note claims, not where it came from. Interview notes tells you nothing in six months. Customers treat setup as a test of trust tells you exactly what is inside and matches the words you would search for when you need it.
This is the single highest-value habit here, because titles are what search sees first and what you scan when choosing between twelve results. Renaming one badly titled note takes ten seconds. Do it whenever a bad title costs you a moment of hesitation.
Maintenance should cost minutes a month, or you have built a second job and called it a system.
Keep the source attached
Record the link, the author, and the date at the moment of capture, along with a page or timestamp where it applies. Adding provenance later is disproportionately expensive, because you have to reconstruct where you were and what you were reading.
For anything you might publish from, add one detail that would let you find the material again if the link dies: the exact title, an issue number, the name of the report. Web pages move constantly, and a note whose source cannot be verified cannot safely be quoted.
Separate quotation, paraphrase, and opinion
Three marks are enough. Quotation marks for the author's exact words, plain text for your paraphrase, and a consistent prefix such as ME: for your own view. Pick a convention and never vary it, because the value comes entirely from being able to trust it at a glance.
This is not pedantry. It is what makes a note safe to draft from at speed, and it is the difference between a citation and an accidental lift. It also preserves something more useful than the source: the record of what you thought when you first read it.
Repair on contact
The cheapest maintenance happens when you already have the note open. If you retrieved it, you are the best-informed person it will ever meet. Fix the title, add the missing link, delete the duplicate paragraph, and move on with your actual work.
This rule concentrates all the effort on material you genuinely use. Notes you never open stay untidy and cost you nothing, which is the correct outcome. A collection does not need to be uniformly clean; it needs to be clean where it is load-bearing.
Run a short monthly reset
Set a timer for twenty minutes once a month and work in a fixed order: empty the inbox, archive projects that have gone quiet, merge obvious duplicates, and repair only the broken links belonging to live work. Stop when the timer goes, even mid-item.
The timer is the important part. Unbounded tidying expands to fill an afternoon and produces nothing you can publish. If the twenty minutes never suffice, the problem is upstream: you are capturing more than you process, and the answer is to capture less.
Decide what to delete
Delete accidental captures, exact duplicates, links you saved without reading, and anything you cannot verify or picture using. Ambiguous material can be archived instead of removed, which costs nothing and keeps search results calmer than leaving everything in the working set.
Do not delete on the grounds that a note is old or that you have changed your mind. Superseded thinking is often the most useful thing in the file, because it records how a view developed, and that record is exactly what a reflective piece needs.
A twenty-minute reset on an eleven-item inbox
Inbox, start of the month: Untitled 4. Read later. Interesting thread. Screenshot 2026-04-02. Notes from Thursday. Untitled 7. Pasted article on remote hiring, full text, no comment. Link to a pricing page, no note. Meeting, no attendees listed. Half a sentence beginning 'the reason nobody does this is'. Duplicate of the note about onboarding trust, saved twice from two devices.
After twenty minutes: five deleted, including both untitled fragments, the screenshot, the unread link, and the pasted article, which was replaced by its single useful paragraph plus the source URL. Two merged, the onboarding duplicates becoming one note with both timestamps kept. Two renamed and filed to live projects: Notes from Thursday became Print supplier lead times are now four weeks, and the meeting note gained its attendees and a decision line. Two left in place, including the half sentence, which is still worth finishing. Timer stopped with the inbox at four items.
Nothing here required a new system. Five deletions, one merge, two renames, and a stopping rule handled the whole backlog with eight minutes to spare.
Common mistakes
| The mistake | Why it costs you, and what to do instead |
|---|---|
| Treating tidying as writing | Reorganising gives immediate visible progress with no risk of failure, which makes it the most attractive way to avoid a difficult draft. Watch when the urge appears. If it arrives on drafting days, it is avoidance rather than maintenance. |
| Scheduling a full review of everything | Reviewing an entire collection is unfinishable and gets abandoned, taking the habit with it. Review only live projects and the inbox. Older material stays reachable by search, which is what search exists for. |
| Fixing links nobody needs | Chasing every dead URL is hours of work for material you may never open again. Repair links in notes attached to current work, and leave the rest broken until a project actually calls one of them back. |
| Renaming into a rigid pattern | Enforced title formats with dates, codes, and prefixes look disciplined and make search worse, because the meaningful words end up buried. Write titles the way you would describe the note to a colleague, and let search do the rest. |
Your checklist
- Rename any note whose title does not state what it claims.
- Add link, author, and date at capture, plus a detail that survives link rot.
- Mark quotation, paraphrase, and your own view with one consistent convention.
- Repair notes you open for real work, and ignore the ones you do not.
- Run a twenty-minute monthly reset in a fixed order and stop on the timer.
- Archive anything ambiguous rather than agonising over deletion.
Key terms
- Provenance
- The record of where a note came from: source link, author, date, and location within the source, kept so a claim can be verified before publication.
- Link rot
- The steady failure of saved URLs as pages move, close, or go behind paywalls, which is why notes need identifying details beyond the address itself.
- Dormant project
- A project folder with no activity for a set period, archived during the monthly reset so the working set reflects only what is genuinely live.
Turning a pile of notes into something?
Frequently asked questions
How often should I review the whole collection?
+
Almost never. A full review is a project in itself and rarely improves any piece of writing. Keep the inbox and live projects current, then trust search and links to surface older material at the moment something actually needs it.
When is a note safe to delete?
+
When it is an accidental capture, an exact duplicate, an unread link, or something you cannot verify or imagine using. If deletion makes you hesitate, archive it. The point is to keep the working set small, not to reduce total storage.
What should I do about dead links?
+
Repair the ones in notes tied to current work by searching the exact title, and record enough detail that the material can be identified without the URL. For everything else, leave it. Most dead links are attached to notes you will never reopen.
Do tags need maintenance too?
+
Only if you have too many. Once a tag list is longer than a screen, it stops being navigation and becomes a second search problem. Merge near-duplicates during the monthly reset and delete any tag used fewer than three times.
Can an assistant clean up my notes?
+
An assistant can propose titles, spot likely duplicates, and draft summaries of long captures, which removes much of the tedium. Approve each change yourself, particularly around attribution, since a merged or summarised note can quietly lose the boundary between a source and your own view.
How do I recover a collection that is already a mess?
+
Do not start at the beginning. Start with the piece you are writing now, repair only the notes it needs, and apply the capture rules to everything new. Within a few projects the material you actually use will be in good order.