The notecard system

If people know one thing about how Robert Greene works, it's the notecards. Ryan Holiday has written about them, made videos about them, and built his own career on a version of them. Billy Oppenheimer, who now researches for Holiday, says questions about the system are "the most common email I get." Greene himself mentions the cards in almost every long interview, usually in a sentence or two, usually with a number attached.

This chapter collects all of it. It starts with what a card looks like, then walks through the system step by step, then goes through each part in detail: the wait, the counts, the colours, the boxes, the move from cards to chapters. After that come the other people's versions (Holiday's, Oppenheimer's, and one researcher's that went wrong), the history of the idea, and Greene's case for doing it by hand. It ends with what this means for an AI version, which chapter 11 designs in full.

The reading that comes before the cards is in chapter 3. What happens to a chapter's cards once the writing starts is in chapter 5. This chapter sits between them.

A warning about the numbers. Greene has given card counts in at least five interviews and written sources over 14 years. They don't all agree, and none of them comes from anyone actually counting a box. The table at the end lists every figure with its source. In the text, I give ranges.

What a card is#

Put together from everything on the record, a Greene card looks like this:

• A 4×6 index card. At the 2023 live event, Greene asked Holiday whether he was "a 3x5 or a 4x6 guy." Holiday said 4×6, and Greene answered, "didn't I tell you to do 4x6?" (Holiday live, 2023). Holiday says the same in his video: "I use 4x6 note cards cuz this is the method that I learned as a research assistant to the great Robert Greene" (Holiday's notecard video).
• Handwritten. "At the end I'll have 2, 3,000 cards… handwritten, handwritten" (Rogan, 2014). Oppenheimer's account says he writes them with a fountain pen (Oppenheimer).
• Coloured by subject. "Each part has the title of the book on it and is color coded, have different colors of cards, depending on the… subject that I'm dealing with" (Mixergy, 2012).
• Labelled with its source. The title of the book, per the same Mixergy quote, and the page: on the Mary Shelley cards he notes "what page it's on" (Perell, 2024).
• About one topic. "If it's a good book, I may have like 10, 20 cards with different topics on it" (Roland).
• Filed by theme. "Notecards organized thematically" (Goins, 2013).

Two details often credited to Greene come from Holiday's description of his own cards, not from Greene: the theme written in the top-right corner, and making a duplicate card when a note fits two themes (Holiday, 2014). Both are likely inherited, since Holiday learned the system from Greene. But Greene himself never describes the corner, and the only first-hand mention of duplicates is Holiday's, at the live event, with Greene listening: "if you want to put one note card in two places you have to do it two times" (Holiday live).

What's on the card itself, in words? Here the record is thin. Greene describes the cards as notes, not transcriptions. He talks about boiling a book down "into like little nuggets of really interesting truths" (Roland), and about taking "all the little parts of that book that fit the theme" and putting them "on a note card" (Perell, 2024). His harshest criticism of a researcher was for printing passages onto cards instead of writing them (more on that below). So a Greene card is probably mostly his own words, with short quotes where the wording matters. I haven't seen a photo of one up close, and he has never read one out in an interview.

The system, step by step#

Here is the whole process as best I can reconstruct it. Each step comes from at least one first-hand source. Where a step depends on Holiday's or Oppenheimer's account rather than Greene's, I say so.

1. Read with a pen. Mark the margins with side lines, underlining, exclamation marks and X's. Take no notes yet. (Chapter 3 covers this.)
2. Put the book aside. Wait somewhere between a few weeks and two months.
3. Go back to the marks. Reopen the book and reread the marked passages. Some of what jumped out the first time no longer does.
4. Write the cards by hand. One topic per card, on a 4×6 card whose colour matches the subject. Note the book and the page.
5. Make copies when a note fits two themes. (Holiday's account.)
6. File the cards by theme in long boxes. The boxes look like shoeboxes.
7. Repeat for every book. A good book gives 10 to 31 cards and a bad one 2 or 3. A project ends with 2,000 to 3,000 cards.
8. Sort the themes into chapters. For the Sublime, "20-some" candidate chapters were whittled to 12.
9. For each chapter, pull its cards. About 60. "Play" with them, write the salient point from each one onto a sheet of paper, and find the chapter's theme in that sheet.
10. Build the opening story from its cards. For the envy chapter, five cards from one Mary Shelley biography, each with its page.
11. Look for gaps. The colours show which kinds of material a chapter is missing.
12. Read more and card it. A second, third or fourth biography at the last minute, carded the same way.
13. Write a schema, then a first draft, from the cards. (Chapter 5 and chapter 6 take it from here.)
14. Start the next book with an empty box. Cards don't carry over.

The rest of the chapter goes through these steps in more detail.

The wait#

Greene doesn't make cards while he reads. He marks the book, puts it down, and comes back later. The most precise statement of the gap is to Olivier Roland:

I like writing in the book etc. And then maybe 2 months later, I go back to that book and I take notes on cards based on my little highlights.

Other accounts give a shorter wait:

• Greene in a Reddit AMA, as quoted in Holiday's 2014 post: "A few weeks later I return to the book, and transfer my scribbles on to note cards" (Holiday, 2014).
• Holiday, for his own version: "A few weeks after finishing the book, return to it and transfer those notes" (same post). In his video he says "sometimes the reason I take a couple weeks before I pick it up or months" is the size of the job (Holiday's video).
• Oppenheimer, for his own version: he waits "a week or two" after finishing a book (Oppenheimer).

So the wait runs from one week to two months, depending on who's describing it and when. The two-month figure is Greene's own and the most recent of his statements. The "few weeks" is also Greene's, from an earlier AMA. Neither is a rule. "Maybe two months" sounds like a description of what tends to happen when you're reading 300 books and carding them in batches, not a timer.

Why wait at all? Greene doesn't say directly. Holiday does, in the video:

The reason you're waiting is that some of the stuff that jumps out at you when you read it the first time, when you go back through it the second time, it's not going to jump out at you.

He adds that the gap "is really important cuz it's the second pass that's now allowing you to engage with the material one additional time" (Holiday's video). In his 2014 post he calls it separating "the wheat from the chaff." Oppenheimer frames the same idea as "time as a filter."

There are two separate effects in that explanation, and it helps to pull them apart.

The wait is a filter. Some passages are exciting in the moment because of what came just before them, or because you were in a certain mood. Weeks later, without that context, they look ordinary. Coming back cold drops them. This is a quality filter that costs nothing except patience.

The wait is also a second reading. By the time he cards a book, Greene has read it once with a pen and then gone back through every marked passage. The material passes through his head at least twice, and the second time it's being judged against everything else he has read since. A passage that seemed minor the first time can look important once three other books have touched the same idea.

There's a third effect that no one mentions but follows from the numbers. Greene reads hundreds of books per project. If carding lags reading by weeks or months, then at any moment he has a backlog of marked books waiting to be carded. When he cards one, the books he's read since are fresh in his mind. So the cards from book 50 get written by someone who has read 70. That is a quiet form of cross-referencing that the system gets for free.

How many cards a book gives#

Greene's figures, in order:

The figures agree well. A bad book gives two or three cards. A good book gives 10 to 31, and the odd exceptional one gives up to about 50. The Reddit figure fits: several dozen books (say 30) at an average of about 10 cards each makes 300.

The precision of "20 to 31" is odd. Thirty-one is not a number people round to. It sounds like a real count: possibly the most he got from one book during the Mastery research, said in a 2012 interview recorded while that book was coming out. I'd read it as "up to about 30, and I remember one that gave 31."

Two of the figures are not per book.

The 50 Leonardo cards are per subject. "If I want to do Leonardo da Vinci, I have 50 notecards that break him down from every possible angle" (Mixergy). For Mastery, Leonardo is a major figure, and those 50 cards would come from several biographies, not one. Compare Darwin, for whom he said he "would have to read three or four books" (Greene's blog). A central figure in a Greene book might be worth 30 to 50 cards from three to five sources.

The five Shelley cards are per theme within a book. He read "a thick biography about Mary Shelley, 500, 600 pages," and "the elements that have to deal with envy in her life are scattered throughout the book." So he made "five note cards that detail every instance of envy in her life, the quality of envy, the nature of envy, what page it's on" (Perell, 2024). That biography probably gave other cards too, on other themes. The five are the envy subset, and they became the opening story of the envy chapter in The Laws of Human Nature.

The Shelley example is the clearest picture anyone has given of what a card is for. It isn't a summary of the book. It's one theme's worth of material pulled out of a book where it was scattered, with page numbers so he can go back. The card system turns a 600-page book organized by time (her life in order) into five cards organized by idea (envy). That reorganization is the point.

Holiday's own counts run higher. At the live event he said that some books are "worth like 60 note cards and you're like, oh, I got to spend like 4 days copying that out" (Holiday live). In his video he says a big book's extracts "might take two or three days of work" (Holiday's video). Oppenheimer runs lower: he picks 5 to 10 cards from a book (Oppenheimer). The three systems sit at different points on the same scale, and that's partly about what each one keeps. Holiday copies long passages. Oppenheimer writes short notes in his own words. Greene sits between them.

How many cards a book project ends with#

The totals:

• "After going through several dozen books, I might have three hundred cards." Greene, Reddit AMA, via Holiday, 2014. A progress figure, partway through a project.
• "At the end I'll have 2, 3,000 cards." Greene, Rogan, 2014. The end-of-project total.
• "A box of two thousand cards that I can sift through with my fingers." Greene on the Knowledge Project, quoted by Oppenheimer.
• 3,000 to 4,000 cards "on average." A Medium post about Greene's method makes this claim. I couldn't open it (the site returned an error) and found no first-hand source for the figure, so I treat it as unverified.

So the end total is 2,000 to 3,000, from Greene twice.

Does that square with the per-book counts? Take 300 books. If two-thirds are weak and give two or three cards, that's about 500 cards. If the remaining hundred average 20, that's another 2,000. Total: around 2,500. The figures hang together, with one assumption that Greene never states: that most books he reads are weak. Given what he says about triage ("a lot of sifting through garbage"), that seems likely.

The 3,000-to-4,000 claim would need either more books or a higher yield. It isn't impossible. But it has no source I could read, and the two first-hand figures both stop at 3,000.

As a sense of scale: a card holds maybe 50 to 100 handwritten words. At 2,500 cards, the box holds something like 125,000 to 250,000 words of notes. The Laws of Human Nature runs to 624 pages. So the box for a Greene book is probably close to the length of the book itself, and possibly longer. The material is reduced from 300 books to one book's worth of notes, and then rewritten into the book.

Greene doesn't give the time all this takes, beyond "excruciating." In 2012 he said "I've spent thousands of hours accumulating information" (Mixergy), which covers the reading and the cards together across his career. Holiday's figure of four days for 60 cards gives a rough rate: about 15 cards a day, if a day means several hours. At that rate, 2,500 cards is around 170 days of carding. Spread across a project of four or five years, that's plausible. It would also explain why he calls the cards the hard part.

"The Bill Bradley part"#

The Rogan passage is worth quoting in full, because it's the only place Greene says how the card work feels (Rogan, 2014):

I read three, 400 books and I read them, I don't skim them, and then I go back and I do these note cards, and it's the note cards that's excruciating. That's like the Bill Bradley part.

The Bill Bradley reference points back to a story he'd told Rogan about 25 minutes earlier, from Mastery. Bradley, the Princeton and Knicks basketball player, was "tall, awkward, slow" as a boy and trained himself with special glasses that stopped him looking down while he dribbled, for "three or four hours" at a stretch, including in the corridors below deck on a ship to England (Rogan). Greene's term for it is "resistance practice": "practicing at what is painful," and learning to enjoy the pain (Rogan).

So when Greene calls the cards "the Bill Bradley part," he's putting them in the same category: the drill. Reading is the game, and he enjoys it. Carding is the practice, repetitive and slow, that makes the game possible. Bradley isn't a card user and isn't part of the system's history. He's Greene's word for its cost.

The colours#

The colour coding is the most cited detail of the system. Greene confirms it in his own words in 2012 ("color coded… different colors of cards," Mixergy). The specific scheme comes from his Reddit AMA, as quoted in Holiday's post. For The 33 Strategies of War:

The quote goes on to explain what the colours are for: "I could glance at the cards for one chapter and see no blue or green cards and realize a problem" (Holiday, 2014).

That's a design decision worth noticing. The themes (the chapter topics) are written on the cards. The colours are a second, independent axis: the domain the material comes from. A chapter on, say, counterattack strategy could be filled entirely with battles. The colours make it visible when that happens. If the cards on the table are all yellow, the chapter has no politics, no art, nothing from outside war. For a book whose whole claim is that the strategies of war apply to daily life, a chapter with only yellow cards is failing at the book's job.

So the colours check range. That ties to something Greene says about his books in general: that they deliberately cover the "full gamut of civilization going back 5 or 6,000 years," and that he picks stories of women and minorities on purpose so the book isn't "just about white men" (Roland, Perell, 2024). A colour scheme can't check that directly, but it's the same instinct: look at the spread of a chapter's sources, not just their quality.

Two limits on what we know:

• The four colours are for one book. Greene says the colours "depend on the subject." Nothing on the record gives the schemes for the other books. A book about human nature might split neuroscience, history and psychology, but that's a guess.
• The scheme is from about 2006. Whether he still colour-codes after the stroke, for the Sublime, isn't stated anywhere. He did still use cards for that book (see below).

Holiday doesn't colour-code his own cards. He mentions instead that his cards show "different colors of ink" because he adds to them over time (Holiday, 2014). His printed cards for the virtue series use a different second axis: each card has the four book titles printed across the top, and he circles the one it belongs to (Holiday live). Greene's colour is "what type of material is this?" Holiday's circle is "which book is this for?"

Page numbers and sources#

A card carries its book's title (Mixergy) and its page (the Shelley cards). That's what makes the cards a research tool rather than a set of notes. When Greene builds the Shelley story, he can go from the card back to the page in the biography and reread the passage in full, in context.

Holiday shows the same habit on camera. He holds up a card that reads, in his description, "12,000 words of comments on one book, page 619, McCullough, John Adams," and says "How did I know that was on page 619 of the McCullough biography? Cuz it's right here on a note card" (Holiday's video). Oppenheimer puts the book title and page numbers on the back of each card, or the podcast title and timestamp, or the article and author (Oppenheimer).

This matters for the rest of this report more than any other detail. A card that points to a page can be checked. A card that doesn't can only be trusted. The printed books mostly don't carry the page references through to the reader (see chapter 3), but the cards do. The whole verification design in chapter 13 rests on keeping that link alive from page to card to sentence.

The boxes#

The cards live in boxes. Greene mentions "shoeboxes" himself, in the 2013 Goins interview (Goins):

I'm an organization freak. I put everything, all my ideas, all my research onto notecards. Notecards organized thematically. And before I've ever written a page, I have shoeboxes full of these elaborate notes.

The best eyewitness account is Holiday's, from when he started as Greene's research assistant (Holiday's video):

I remember I just started working for him and he takes me into his office and he starts pulling out these boxes. They look like long shoe boxes, and he says, "You want to see how I wrote the 48 Laws of Power?"

Holiday goes on: "It's organized in all these different… He's like, 'It's all about the note cards. The note cards are the building blocks'" (Holiday's video).

That scene tells us a few things. Greene kept the 48 Laws boxes for years after the book came out. Greene told the 2023 audience that Holiday had contacted him "some 16 years ago," so Holiday started around 2007, and the boxes were about a decade old. They were organized by theme, with dividers. And the cards were the thing Greene chose to show a new assistant first.

The boxes are physical constraints too. Oppenheimer's box holds about 1,500 cards (Oppenheimer). At that size, Greene's 2,000 to 3,000 cards would fill two boxes, which fits the plural in every account. Holiday uses Cropper Hoppers, plastic cases sold for storing photos, and says "this is thousands of note cards" of one of them (Holiday's video).

From cards to chapters#

The cards are the raw material for the book's structure, not just its content. Greene said so to Rogan: "I can take these cards and move them around and the book is created out of the cards" (Rogan). And to Perell, describing the Sublime in 2024 (Perell, 2024):

When you have the note cards, you organize them into themes, and those themes become your chapters. And I started off with maybe 20-some chapters for this book. They're now 12 because I whittle them down.

So the chapter list comes out of the cards, bottom up. The themes he files the cards under during the research become candidate chapters. Some merge, some get dropped, and the 20-odd themes end up as 12 chapters. This is the opposite of an outline-first method like Robert Caro's, where the book is boiled down to a few paragraphs before the research starts. Chapter 9 compares the two.

Then, per chapter, the cards come out again. In the same Perell passage, Greene describes the gap between chapters: "When this chapter is over… it's like all the air goes out of a balloon. I finally relax and now I start gearing up for the next chapter and I go through my note cards and I have fun and I play with my note cards." Then: "I go through all the cards and I note down salient things in each one on a paper. And from that I see themes that'll turn into the chapter" (Perell, 2024).

How many cards is that? "When I get down to write a chapter, I'll have like 60 cards. And then I can look at them and I can say, 'Oh, this is what I'm going to write about.' It's all organized already" (Roland).

There's a puzzle in those numbers. Sixty cards per chapter times 12 chapters is 720. Times 18 chapters (The Laws of Human Nature) it's about 1,080. Neither reaches 2,000 to 3,000. Three explanations fit, and they probably all apply:

1. Not every card gets used. Weak cards, cards on dropped themes, and cards that duplicate others stay in the box. A book that started with "20-some" chapter themes and ended with 12 left eight themes' worth of cards unused.
2. Duplicates. If a card that fits two themes is copied, the total count includes both copies, but each chapter only sees one.
3. "Like 60" is a rough figure. It was said to make a point about how organized the material is, not as a count.

There's also a second layer that the 60 doesn't include. After he picks the opening story, Greene often reads more: "At the last minute I'll read a second, a third, even a fourth biography" (Perell, 2024). Those late books presumably get carded too, and added to the chapter's pile.

The step from cards to draft has its own stages: a sheet of salient points, a theme, a story, a "schema" of main points, then a first draft that takes about two months. That's chapter 5's subject.

The cards don't carry over#

One detail sets Greene's system apart from Holiday's, and from most note-taking advice. His cards are made for one book and stay with it. At the live event, after Holiday explained that he runs four boxes for four books at once and moves unused cards between them, Greene answered (Holiday live):

You take only like a year or two to write a book. I take like five years or so, and you know, I know that there's something in there that will be good for my next book, but by the time I get to my next book so many years have passed that it's just useless for me.

So the box is a project tool, not a lifelong archive. Greene's books change subject every time (power, seduction, war, fearlessness, mastery, human nature, the sublime), and he takes years over each one. By the time he starts the next, he says, he's a different reader and the old cards don't speak to him.

That doesn't mean old material never reappears. He told Rogan in 2014 that he'd read Phil Jackson's latest book and was "going to include him in my next book" (Rogan). He has also said "I have no idea where I've pulled that story from" about an Edward Bernays anecdote from years earlier (Williamson, 2021), which suggests stories travel in his memory even when the cards don't. And he kept the 48 Laws boxes long enough to show Holiday. Kept, then, but not reused.

Holiday's system is the opposite. "I have tens of thousands of them for over 20 years," he says, organized into commonplace books per project and per subject, including "a commonplace book of just life advice" (Holiday's video). He says the value came from accumulation: "I got some advantage out of it when I started on my first book, but it was my second and my third" (Holiday's video). Holiday writes shorter books faster, on related subjects (Stoicism, then the four Stoic virtues), so cards from one book often fit the next. Greene's books don't share material that way.

Where the system came from#

Greene tells two origin stories, and he tells them in the same breath. To Perell in 2024:

I discovered when I was doing the 48 laws, I did a lot of research for that book in a very condensed period of time, that the material was overwhelming me. You know, I was reading so many books from all different cultures, periods. How am I going to organize this? I can't take notes on paper. So I developed a note card system.

And then, immediately: "And I had been using this system since I was a child. My mother showed me in one of my drawers in our house that I used to have… I used to take notes on cards and organize. So, it's ingrained in me for some reason" (Perell, 2024).

These fit together if you read "developed" as "rebuilt for scale." He had a childhood habit of putting notes on cards. When the 48 Laws research overwhelmed him, he turned that habit into a system with themes, colours and boxes. In 2013 he told Goins "my process is something I've developed over the years" (Goins), which supports the gradual version.

The 48 Laws came out in 1998, after about two years of work. So the system as Holiday saw it dates from roughly 1995 to 1997, and had been in use for about a decade when Holiday joined.

Then Greene adds a line about where else the idea exists: "I know the writer Vladimir Nabokov used a note card system. I know Napoleon had a note card system. I didn't invent it, but I wasn't… nobody taught me that system. I came upon it on my own, which a lot of people have done" (Perell, 2024). The lineage section below checks those claims.

Holiday's account#

Ryan Holiday is the main outside witness to the system. He describes himself as having been "a research assistant under the great Robert Greene" (Holiday's video). A 2016 Reddit comment by Holiday says he worked on The 50th Law and Mastery, and a blog recap describes the job as "transcribing interviews and reading books he (Greene) didn't want to read" (a quoted Reddit comment; The Waiter's Pad). At the live event Holiday mentioned transcribing "hundreds of hours" of 50 Cent interviews (Holiday live).

Greene's description of the arrangement in 2012 is worth noting for what it says about the cards. "He's doing some of the research… it's not like I'm getting slave labor… I'm saving the best books for myself… He's learned about my notecard system. I didn't even have to tell him that" (Mixergy). So even with an assistant, Greene kept the best books for himself. The assistant read the rest. The cards from the important books were Greene's own.

Holiday's key reports about Greene's system:

• The shoeboxes and the 48 Laws tour. Quoted above.
• "It's all about the note cards. The note cards are the building blocks." Greene's maxim, as Holiday remembers it.
• The 4×6 size. Confirmed by Greene at the live event.
• No digital version. The exchange that has become the system's motto:

I remember I asked Robert Greene, I was like, "Can I just do this all digitally?" And he was like, "No, the whole point is that it's physical, not just the painful part of doing it."

That's from Holiday's video. Notice the second clause. Greene's objection, as Holiday reports it, is not only that the pain does you good. Something about the cards being physical objects matters in itself. The section on why physical comes back to this.

Holiday describes his version as "a perverted version of a system taught to me by the genius Robert Greene," and Greene's as "neat, clean and orderly" (Holiday, 2014). The differences:

The biggest differences are about copying and about time. Holiday copies more text, and he keeps everything forever. Greene writes shorter notes and starts fresh each time.

The typing confession#

At the live event Holiday admitted to a habit that strays furthest from Greene's method (Holiday live):

Sometimes I type the passages… and I don't feel like it's cheating. It's still flowing through. I'm doing it because my handwriting, I can't get it all on the card. It's the only way I can shrink the font down… I print it out and then I cut it out with scissors and then I tape it, or I get one of my kids' glue sticks.

His other reason: "my handwriting is so bad I found that I've been introducing errors and then I have to go back to the text." Greene's response: "I forgive you. It's all right. It's not a sin" (Holiday live).

The error point is the more interesting of the two. Hand copying introduces transcription mistakes. For a writer whose books quote other books, a mistake on a card becomes a misquote in print. Typing the passage and pasting it removes that class of error. It also keeps the effort Greene cares about, since Holiday still chooses the passage, types it, prints it, cuts it and glues it. Greene's "it's not a sin" accepts that. What separates Holiday's typing from the researcher Greene fired is the next section.

The researcher who printed passages#

In the same exchange, just before Holiday's confession, Greene told the story of his worst researcher (Holiday live):

You were the best researcher I've ever had, by far. I had the worst researcher a couple years ago… I told him my note card system. What he did is he literally digitally copied passages from the books, and I don't know how he did it, he printed them on the note cards. He didn't take the time to write and to think about what they might be. He just copied passages on note cards.

And the punchline: "Well, I could just read the whole book through instead of going through your stupid note cards" (Holiday live).

Put Holiday's habit and the researcher's next to each other and the difference is clear. Both put printed text on cards. Holiday chose each passage, decided where it went, and wrote his comments around it. The researcher appears to have pulled text in bulk without that judgment. Greene's complaint isn't about the printer. It's that a card full of the book's own words, picked without a theme in mind, carries no information the book doesn't. If the cards aren't a reduction, he might as well reread the book.

This lines up with what Greene says about his other researchers. He hired UCLA grad students and USC PhD students and "had no luck": "they're not able to sit with boredom like you need to… They have an ego. They think they know better than me. They don't listen to the process that I try and tell them" (Perell, 2024). The printed-passage researcher is a case of not listening to the process. At the same event, an audience member brought up Oppenheimer's writing about Greene's assistants not grasping "the idea of being interested in boredom," and Greene answered: "Boredom is a great thing in life… you have to develop internal skills for dealing with it" (Holiday live).

The worst researcher's mistake is the main design warning for any AI version. A model can pull every marked passage from a scanned book in seconds. If that's all it does, it has reproduced the researcher Greene let go. The section at the end returns to this.

Holiday's "cons are pros"#

Holiday gave the clearest statement of why the system's costs are its point, at the same live event, right after Greene finished talking about AI (Holiday live):

The thing about the note card system is that all of the cons are actually pros, right? The fact that it is hard, that you can't scale it, that it takes forever, that your hand aches, that if you want to put one note card in two places you have to do it two times… that it's physical, that you can lose it, all of that is actually why it works.

He continues: "if you want to get it from the book to the note card… you have to physically take it from here, it has to flow from your mind through… and that it's hard is the reason that it works." The things digital tools allow, highlighting and passing text around without limit, "are advantages in the sense that they remove the difficulty, but by removing the difficulty they remove the point. They remove all the value" (Holiday live).

In his own video he puts it in four words: "The inefficiency is the efficiency." And: "I want you to sweat for it. I want it to take time. I want it to be difficult" (Holiday's video).

The argument has a specific mechanism, and it's worth stating it plainly, because "hard things are good" is not an argument. The claim is that each cost forces a decision:

• Copying by hand is slow, so you only copy what's worth copying. The cost filters.
• Duplicating a card means writing it again, so you only put a note under two themes if it truly belongs there. And writing it again is one more pass through the material.
• Moving cards around is physical, so reorganizing the box means handling every card, which means rereading it.
• You can't search the box, so you have to remember roughly where things are, which means knowing the material.

Each of those is a point where a digital system would save time and the person would skip a judgment. That's the testable part of the claim: not that pain helps, but that friction placed at each step forces a decision a frictionless tool lets you skip. It's also exactly what a pipeline designer has to think about. Chapter 9 looks at the argument as part of Greene's wider case against shortcuts.

Then Holiday handed the floor back to Greene, who quoted the writer the caption calls "Douglas Hofer," almost certainly Douglas Hofstadter: training for weeks to climb Everest versus taking a helicopter to the top for the same view (Holiday live).

Holiday's variant, in practice#

Holiday has refined his own version over 20 years, and his descriptions show where a system built on Greene's can go.

Printed cards. "I got like 10,000 note cards printed from a printer down the street from the bookstore and it says courage, discipline, justice, wisdom at the top of the note card, and when I'm doing the note cards I circle which book that note card is going to pertain to" (Holiday live). That was for his four-book series on the Stoic virtues. "I have four boxes for four books… some of the stuff that I didn't use in the first book moves over to the second, which moves over to the third" (Holiday live).

Commonplace books. He calls each box a "commonplace book," after the old practice of keeping a notebook of quotes and observations. He shows his box for Wisdom Takes Work and pulls cards from it on camera: a story about Richard Feynman from page 325 of Leonard Mlodinow's book, with his own riff underneath (Holiday's video).

Reshuffling. "Having it laid out allows me to organize it, put it in themes, but move them around. I go, 'Actually, no, this doesn't go here, this goes here. This belongs in two categories.' And now I've got to write it again" (Holiday's video). That's the cards-to-chapters step Greene describes, from the other side.

Counts. "Thousands" of cards for The Obstacle Is the Way (Holiday, 2014). "Tens of thousands" over 20-plus years (Holiday's video). Many of his cards hold only a word or two, he says in the 2014 post.

Kindle. He says taking notes on a Kindle is "too easy and that's the problem," and calls the physical side "irreplaceable" (Holiday, 2014).

Billy Oppenheimer's account#

Oppenheimer works as Holiday's researcher, the job Holiday once did for Greene. His write-up of the notecard system is the most detailed public description of any version (Oppenheimer). Most of it is his own practice, but it includes the only published Greene quotes about why the cards are physical, taken from Greene's appearance on the Knowledge Project podcast.

The Greene quotes, as Oppenheimer gives them:

• "Writing things out by hand has a logic to it… [it] links closer and faster to the way my brain works."
• He thinks "more deeply than when typing on a computer."
• He wants "a box of two thousand cards that I can sift through with my fingers… I can't do that on a computer. It's not the same process."

Oppenheimer's own system:

• No copy-and-paste. He writes every note by hand to add "mechanical resistance," so only material worth the effort gets in.
• A week or two of waiting, then 5 to 10 cards per book.
• Write for a stranger. Each card should make sense on its own, with enough context for a future reader who has forgotten the book. On the back: the title and page numbers, or the podcast and timestamp.
• A title at the top of each card for scanning.
• A "Waiting Room." New cards that don't fit an existing section go into a holding area. When three or four related cards turn up, they become a new section with an index card in front.
• Weekly review before his Sunday newsletter, and near-daily use.
• Mix of sources: about 75% books, 13% podcasts, 10% articles, 2% videos.
• A box of about 1,500 cards, about a quarter full at the time of writing.

Two of these depart from Greene in ways worth noting. The Waiting Room is a bottom-up way to grow new themes, which is close to how Greene's 20-odd themes emerge, but made explicit. And "write for a stranger" is a standard that Greene's notes may or may not meet: his cards only have to make sense to him, within one project, for a few years.

Oppenheimer also says he doesn't make cards for Holiday. His job is to find material for Holiday's system, not to fill it. That keeps the rule that the person who writes the book writes the cards, which is the same rule Greene's printed-passage researcher broke.

The lineage#

Greene says he came to the system on his own, and also that others had it before him. Here's what I could check.

Niklas Luhmann#

The most famous card system belonged to the German sociologist Niklas Luhmann. Starting around 1952 to 1953, he built a Zettelkasten ("slip box") of about 90,000 index cards, each with a unique number in a branching scheme so that a card could be linked to any other and new cards could be inserted between old ones. He credited it with his output of roughly 50 books and 550 articles. The cards were digitized and put online in 2019 (Wikipedia).

Holiday says his own system "looks very similar" to Luhmann's, though Luhmann's discipline "seems to exceed mine" (Holiday, 2014). He also mentions the German word in his video (Holiday's video). Greene never mentions Luhmann in the interviews I read.

The two systems differ in the way that matters most:

Greene's system is closer to the older tradition of excerpting for a single work than to Luhmann's permanent network. The cards are built to be used up.

Commonplace books and earlier#

Holiday places the system in the tradition of the commonplace book: notebooks where readers copied quotes, observations and anecdotes they wanted to keep. He mentions ars excerpendi, "the art of excerpting," as a skill taught since the Greeks and Romans, and names Montaigne, Emerson, John Adams and Anne Frank among those who kept them (Holiday's video). The move from bound notebooks to loose slips that can be rearranged is usually traced to the 16th-century Swiss scholar Conrad Gessner, who devised a method in which "individual notes could be rearranged at any time" (Wikipedia).

That move, from a notebook to loose cards, is what makes Greene's system work. A notebook keeps notes in the order you read. Cards let you re-sort them by theme, which is how five scattered moments in Mary Shelley's life end up side by side on the table.

Nabokov#

Greene's claim about Nabokov holds up. Nabokov routinely composed on index cards, and they were the basis of many of his works, including Lolita and Pale Fire. His unfinished last novel, The Original of Laura, survives as 138 handwritten index cards, which his son Dmitri published in 2009 as removable facsimiles beside a printed transcription (Wikipedia).

But Nabokov used cards for a different job. He wrote the novel itself on them, in sequence, so he could rearrange scenes. Greene uses cards for research notes and writes the book elsewhere (on a laptop before the stroke, in notebooks after). Nabokov's cards are drafts. Greene's are sources.

Napoleon#

Greene says "I know Napoleon had a note card system." I found no source for this in the material for this report, and I didn't search the Napoleon literature for it. It may be a memory from the 4,000 pages he read for The 33 Strategies of War. It fits his reading of Napoleon as a genius of "organizing information" (chapter 3). It also fits the "Napoleon Bonaparte's genius was organizational" line he gave Goins. But until someone finds the source, I'd treat it as Greene's claim, not a fact. One small coincidence: Oppenheimer's own card box includes a note on Andrew Roberts's description of Napoleon's "extraordinary capacity for compartmentalizing his mind," which he later used in a piece on attention (Oppenheimer).

Others who used cards#

Holiday's video adds a few more: Ronald Reagan, who kept anecdotes and quotes on 3×5 cards in photo binders, which Holiday asked to see at the Reagan Library; Joan Didion, who organized screenplays on note cards; and George Patton, who marked the books he finished (Holiday's video). Robert Caro keeps index cards on a corkboard alongside his wall-sized outline (Every). None of these connect to Greene directly. They show that the 4×6 card is a common tool for people who write from large amounts of material.

Why physical#

Greene's case for doing it by hand comes in three layers, from three different times.

First, the cards are objects you can move. "I can take these cards and move them around and the book is created out of the cards" (Rogan, 2014). And from the Knowledge Project: "a box of two thousand cards that I can sift through with my fingers… I can't do that on a computer. It's not the same process" (via Oppenheimer). This is about spatial work: spreading a chapter's cards on a table, seeing the colours, moving one from here to there. A screen shows a handful of notes at once. A table can show 60.

Second, handwriting changes the thinking. "Writing things out by hand has a logic to it… [it] links closer and faster to the way my brain works" (Knowledge Project, via Oppenheimer). After the stroke, when he wrote the whole Sublime by hand, he said the same about drafting: "when you're handwriting, your thinking really has to slow down. You have to focus very deeply because you can't quickly change things" (Huberman, 2026).

Third, the hand is linked to the brain. This is the most recent and most physiological version. Speaking to Andrew Huberman, a neuroscientist, in 2026 (Huberman, 2026):

The relationship between the brain and the hand is a very, very powerful one. It has very ancient roots. So many of the neural pathways involve the hand and the brain because that's where our intelligence evolved, picking up things and manipulating them and creating tools… So when you handwrite, you're using the brain differently than when you type and look at a screen.

He says that when he tried dictating first drafts after the stroke, "something was lost… my hand literally has to be involved in the creation of it." That's about drafting, not cards, but it's the same belief.

How much of this is supported by research is outside this chapter's scope, and Greene doesn't cite any. What matters for this report is that his reasons are specific. He isn't saying "the old way is better." He's saying three things that could each be tested: that a spread of physical cards supports an overview a screen doesn't; that handwriting slows thinking in a useful way; and that the effort of copying makes the material stick. The first is about the interface. The second and third are about what happens in the person.

That distinction matters for an AI version, because a model doesn't have a hand. It can be given the first benefit (show it all 60 cards at once, or all 3,000). It can't be given the second and third, because those benefits accrue to the person doing the copying. If the model does the copying, the human loses them.

After the stroke#

Did the system survive Greene's 2018 stroke? Yes, at least in part. In 2024, writing the Sublime by hand with one working hand, he described going "through my note cards" between chapters and "constructing the story from my note cards" (Perell, 2024). He still had about 60 cards per chapter, still sorted them into themes, and the chapter list still came out of them.

What isn't on the record: whether he still colour-codes, whether he still writes the cards himself or has help, and how many cards the Sublime produced. The reading for that book was done mostly from his desk and backyard (see chapter 8), so the card-making would have happened at the same desk. Greene's right hand works, and he wrote more than a hundred notebooks with it for the drafts. There's no reason to think he stopped writing cards. There's also no direct statement that he didn't.

What this means for an AI version#

The short post sketched a pipeline in which models do the research assistant's job: reading the books, marking passages, and making cards, while a human picks stories and writes. Chapter 11 turns that into a full design. Here are the parts of the notecard system that design has to deal with, and what each one turns into.

The card itself is easy to represent. A 4×6 card with a book, a page, a theme, a colour and a note maps onto a small record. The pipeline chapter defines it as a database row with a verbatim quote, a separate paraphrase field, the printed page, a theme and a domain. Greene's colour scheme becomes a domain tag, and the "see no blue cards" check becomes a script that counts each chapter's cards by domain and flags gaps. That part of the system translates almost directly, and it's the part a computer does better than a table, since it can check every chapter at once.

Page numbers are the part to protect. Greene's cards carry pages so he can go back. In a pipeline, the page is what lets a script check the card's quote against the scanned page before the card is saved. That's the main defence against invented quotes, which chapter 13 shows are the most common failure in AI research tools. Greene's own system has no such check: a card is as accurate as his handwriting, which is the problem Holiday's typing habit solves. A pipeline can do better than the original here.

The wait becomes a second pass. There's no point making a model wait two months. But the two effects of the wait can be reproduced: a fresh context with no memory of the first read (the filter), and a second pass that judges each mark against everything read since (the cross-reference). The pipeline design runs its second pass with a fresh context against the current theme map for that reason.

Duplicates and reshuffling become free, and that's a loss. In a database, a card can carry two themes at no cost, and moving it between chapters is a field update. Holiday's argument says that's exactly what removes the value: each copy and each move was a forced reread. A pipeline can't give the human that friction back. What it can do is make sure the human still makes the decisions the friction used to force, by putting them at gates.

The printed-passage trap is the main risk. A model that extracts every marked passage onto a card is Greene's worst researcher, at scale. Greene's complaint was that such cards carry nothing the book doesn't. The pipeline's answer, in chapter 11, is to keep the verbatim quote (so it can be checked) and require a separate note explaining why the passage matters to its theme (so there's a judgment to inspect). The test of whether that works is the one Greene applied: would you rather read the cards or the book? If the cards don't save the writer from rereading the book, the stage has failed.

The hand can't be automated. Greene's second and third reasons for working by hand (slower thinking, material that sticks) are benefits to the person, not the box. If a model makes the cards, the writer doesn't get them. The honest version of an AI notecard system says so up front: it trades the writer's absorption of the material for speed and checkability. Whether that trade is worth it depends on whether the writer still reads the key books themselves, the way Greene kept "the best books for myself" when Holiday was doing the rest.

Cards per book is a tunable. Greene's 2 or 3 for a bad book and up to 31 for a good one is a rate of reduction, and a model can be asked for any rate. The design question is what to aim for. Too many cards and the box becomes the book again. Too few and the chapter lacks material. Greene's end total of 2,000 to 3,000 for a 500-page book, and about 60 per chapter, give a starting target.

Every number, with its source#

Sources#