Where we start
This course runs on one collection. Daybreak, a coffee subscription company, employs somebody who has been taking notes for eighteen months: meeting notes, supplier calls, saved articles, half-formed ideas, two live projects. Forty-eight notes. Every session does something different to that same collection, and in session 4 a bench asks it sixteen real questions and counts how many it can answer.
★ What you walk out with tonight. An honest count of your own collection, a sample of ten notes scored on whether you could get them back, and a two-sentence answer to "does any of this actually work?" that names what is evidenced and what is not.
Three neighbours, so you land in the right place. Atomic notes, permanent addresses and linking by idea rather than filing by folder is the Zettelkasten method, and that is learn-obsidian-with-phoebe. Reading technique and turning a book into notes is learn-reading-with-phoebe. Building the thing in a specific tool is the Notion and Obsidian courses. Here the subject is the loop itself - capture, organize, retrieve, review - in whatever you already use.
Saving is cheap, finding is not
5 min liveEvery tool built in the last fifteen years has optimised the save. Clip it, share-sheet it, forward it, screenshot it. Nothing optimised the getting-back, because the getting-back happens eight months later when the tool is not watching. The result is an asymmetry you can draw.
LiveFour things that go wrong, and none of them is laziness3 min▶
People assume the problem is discipline. It is nearly always one of these four, and all four are design problems.
- The note has no title anybody could search for. "Notes", "Untitled", "Interesting". You knew what it was on the day. Search works on words that are actually in the note's name, and there are none.
- The place it lives is too big to look in. One folder with everything in it is not a place, it is a list. Past a dozen or so items you stop scanning and start scrolling, and scrolling does not find things.
- Nothing points at it. A note that nothing links to and nothing sits beside can only be reached by search, so the first problem becomes fatal.
- It was never going to be reread. Some of what you save is saved to end the discomfort of not having saved it. That is a real behaviour with a name, and part 2 has the measurements.
Session 4 turns exactly these four into numbers over a real collection. Tonight is about seeing them in your own.
Self-studyWhat the research calls the pile4 min▶
Accumulating digital material you never revisit has been studied, and it has an instrument. Neave, Briggs, McKellar and Sillence (2019) built a Digital Hoarding Questionnaire across two samples: 424 adults for the structure, then a fresh 203 to validate it. It came out as ten items in two factors, difficulty deleting (six items) and accumulating (four), and the two correlate strongly at r = .736. Both sub-scales correlated at r = .416 with the sheer number of files a person kept at work. Emails were the most commonly hoarded thing in both samples.
A separate qualitative study is worth knowing because it refuses the pathology framing. Vitale, Janzen and McGrenere (2018) interviewed 23 people and found a spectrum running from hoarding at one end to deliberate minimalism at the other, with most people somewhere in between and not distressed about it. So the honest framing is not "you are a hoarder". It is that keeping is frictionless, discarding is effortful, and a collection with no discard step only ever goes one direction.
The Daybreak collection has six notes called some version of nothing: "Notes", "Untitled", "Interesting", "Read later", "Misc", "Ideas". Two of them turn out to matter: one is the meeting where the October launch date was agreed, and one is a competitor tip somebody sent. In session 4, the bench asks for both. One of them is recoverable in the best filing the course builds. The other is lost in every single one, and it stays lost, because nothing points at it and its title is a word.
The study everybody assumes exists
6 min liveThere is no controlled study showing that PARA, a "second brain", or any packaged personal knowledge management method improves what you retrieve, produce or learn. Not a weak one, not a contested one. It was searched for directly, from several angles, and what comes back is blog posts and course pages. That absence is the honest starting point, and it is not the end of the argument, because the mechanisms these methods borrow are among the best-evidenced findings in learning research.
LiveThe finding that should change how you review4 min▶
Roediger and Karpicke gave 120 students prose passages to learn, then either let them restudy or made them try to recall with no feedback. Five minutes later, restudying won: 81% against 75%. That is the result everyone's instincts predict, and it is why people reread.
Two days later it had reversed: 68% for the tested group, 54% for restudy. At one week, 56% against 42%, and the tested group after a week remembered as much as the restudy group had after two days. Their second experiment made it sharper. Studying four times beat everything at five minutes (83%) and came last at one week (40%), while the group that tested three times came last early (71%) and first late (61%).
Then the part that matters most for a note system: the students who studied four times were the most confident they would remember it in a week, and they remembered the least. Rereading feels like learning. Retrieving feels like failing, and it is the one that works. A meta-analysis of 159 effect sizes across 61 studies puts the overall benefit at g = 0.50, medium and reliable, with the caveat that published studies show a bigger effect than unpublished ones, so some inflation is likely.
The application is session 5, and it is not flashcards. It is that a review where you look at your notes is worth much less than one where you try to say what is in them first.
Self-studyHow to quote the offloading study without overclaiming4 min▶
Somebody will tell you that saving things makes you forget them, and cite Sparrow, Liu and Wegner's 2011 "Google Effects on Memory" in Science. Here is what that paper actually did. In one experiment, people who had just been asked hard trivia questions were slower to name the colour of computer-related words (712ms against 591ms) than general words, which suggests a knowledge gap makes you think of a computer. In another, people who typed trivia statements and were told the statements would be erased recalled more of them (0.31) than people told they would be saved (0.22). Being told to try hard to remember made no significant difference; expecting the thing to be available later did.
What that licenses is narrow and worth saying precisely: in a small lab study, with decontextualised trivia sentences typed into a computer, believing a sentence was saved reduced immediate recall of that sentence. What it does not license is any claim about your project notes, about long-term memory in knowledge work, or about a filing method making anyone worse at their job. The task was not remotely like retrieving your own notes eight months later, and the paper does not claim it was.
There is a real tension in this subject, and it is not that one. It is that the most quoted capture rule in the field, Tiago Forte's "keep only what resonates", is deliberately intuitive and non-analytical, while the digital hoarding work measures "difficulty deleting" as a trait that correlates with accumulating more than you can use. Session 2 does not resolve that tension. It gives you a capture rule you can defend and tells you which side it errs on.
What a system can promise, and what it cannot
4 min livePut the two halves together and the claim a note system can honestly make is narrow, specific, and still worth the effort. The oversold version is what makes people quit in month three.
| The claim you will hear | What the sources actually support |
|---|---|
| "Build a second brain and you will never lose an idea" | No controlled study of the method exists. Finding depends on how a note is titled and what points at it, which session 4 measures directly |
| "Set up this app and you have a second brain" | Forte calls this myth number one about his own work: it is "a distributed system made up of apps, devices, websites... and people", not a tool you install |
| "You have to implement the whole system properly first" | His myth number two: the system is "adaptive", and "any incremental improvement immediately makes a difference" |
| "People only ever reread 10% of their notes" | No study found. The number circulating traces to one blogger testing their own workflow, with no participants and no method |
| "Tagging everything makes it findable" | The personal-information-management literature reports no clean winner between folders and tags, and no measured threshold at which tagging starts to hurt |
Audit your own collection
★ 22 min · everyone opens their notesYour real notes, not an example. The rule for the next twenty minutes: no reorganising anything. Tonight is measurement, and measuring before changing is the only way you will know later whether any of this worked.
Count it. How many notes, files or saved items are in the place you keep things? If there are several places, count each and add them. Most people are surprised twice: first by the total, then by how many places there are.
Sample ten. Not your ten favourites. Scroll to a random point and take ten in a row, then another random point if you ran out. Favourites are the ones you can already find, which is the wrong sample.
Ask the three questions of each of the ten. Could somebody search for this by its title alone? Does anything link to it or sit beside it in a place small enough to scan? And, honestly, was it ever going to be reread?
Score the ten. Two yeses or more is findable. One is fragile. Zero is lost, and you are looking at it right now, which is the only time that will ever happen.
Write the number down where you will find it in six weeks. "Lost 6 of 10, 2026-09-20." That sentence is your baseline.
Count the places first, not the notes. Notes app, docs folder, saved messages, bookmarks, the inbox you use as storage, the phone photos of whiteboards. Most people have four to six.
Do not tidy while you count. The urge is strong and it destroys the measurement. Anything you fix tonight you cannot count tomorrow.
Read only the titles for the first pass. If you have to open a note to know what it is, that is the finding, and it has already answered question one.
Be honest about question three. "Was this ever going to be reread?" is uncomfortable and it is the most useful of the three. A collection where most answers are no does not have a filing problem.
Keep the ten. Session 2 rewrites their titles, session 3 gives them homes, and session 5 puts them in a review. The same ten, all course.
Bring the number to session 2. Lost out of ten, and the number of places. Both get used.
If you take one thing from tonight. A collection is not a set of things you have. It is the set of things you can get back, and those two numbers are much further apart than anybody expects before they count. Everything that follows is closing that gap.
Before session 2
25 min- Finish the audit if the twenty minutes ran out, and write the baseline sentence somewhere you will find it: how many notes, how many places, how many of ten were lost.
- Try to retrieve five things you know you saved over the last year, by memory, without browsing. Time yourself. Note which ones you found by searching and which by remembering where you put them; that split is what session 3 is about.
- Find your worst title. The note whose name tells you least about what is inside. Do not rename it yet; session 2 does that, and you want to feel the before.
- Write down one thing you looked for this year and never found. If you can remember what it was about but not where it went, it is the problem in one sentence.
Three questions
Clear all three and session 1 gets stamped in your passport.
1 · A colleague says PARA is "proven to work". What is the accurate reply?
The search was made directly, from several angles, and returned blog posts and course pages rather than outcome studies. A separate academic PKM literature exists but is survey-based and correlational and does not test these methods. Saying "the mechanism is evidenced, the packaging is not" is both honest and still a good reason to do it.
2 · In Roediger and Karpicke's study, restudying beat testing after five minutes and lost badly after a week. Which detail matters most for how you review notes?
The metacognitive reversal is the practical finding. Studying four times gave the highest five-minute score (83%) and the lowest one-week score (40%), with the highest confidence attached. A review built on what feels productive optimises for the wrong thing, which is why session 5 makes you say what is in a note before opening it.
3 · Somebody cites the "Google effects on memory" study to argue that saving notes makes you forget. What does that study actually support?
The task was decontextualised trivia, the samples were small, and the measure was immediate recall of the typed sentence. The paper does not claim more than that. Quoting it as evidence against note-taking stretches it past what it measured, which is the same error as quoting PARA as proven.
What this session teaches, and where it came from
Every fact, its verification tier and the two summariser fabrications caught during research are in the source map.
Session 1 cheat sheet · pin this
You have a baseline and ten sampled notes, and probably a worse number than you expected. Session 2 starts at the front of the loop: a capture rule you can apply in about four seconds without an argument with yourself, and the thing that decides whether a note is ever findable again, which is not where you put it but what you called it.