The best way to acquire knowledge from readings

An effective method for acquiring, retaining, and applying knowledge.

The best way to acquire knowledge from readings


The purpose of visualizing notes is to gain a deep understanding of what you've learned. If all of your notes are very long and you don't break down the knowledge into smaller parts, the understanding you can gain from visualization will be very limited. Real deep understanding doesn't come from the "relationship between two books" but from the "relationship between all the concepts in these two books."

You can only gain a deep understanding of the topics you care about through visual note-taking when you atomize your notes. Atomic note-taking does not mean you cannot have long notes. It means that each concept card should only contain one concept and be supported by its content. To ensure clarity, you should always describe the concept in one sentence and use that sentence as the title of the concept card.


I read a lot. For me, the most frustrating issue with reading is that great books often contain a high volume of content that is not easy to fully digest and integrate with my existing knowledge. Even when I fully digest the content, I often find it difficult to recall knowledge that I learned in the past purely from memory, which makes it hard to apply what I've learned to my current work.

I have met many people who face the same problem in learning, and this issue is what many note-taking (a.k.a. knowledge management) tools are trying to solve. Unfortunately, most of these tools put too much focus on how you store your notes (e.g., folders, graphs, relational databases, etc.) and neglect important aspects of learning such as acquisition (making sense of knowledge), retention (recalling past knowledge), and application (applying knowledge in a real-world scenario).

In this article, I will share a real-world example of how I developed an effective method for acquiring, retaining, and applying knowledge using Heptabase, a tool my team and I created for learning and research. While this is not the only workflow for learning, it has worked extremely well for me, and I believe almost anyone can apply it to their learning.

Method Overview

The famous Feynman learning technique suggests that the best way to develop a deep understanding of a topic is to teach it to a child. I would say that whenever you want to teach something, you need to first figure out the structure of the knowledge and be able to articulate that structure clearly. The method that I developed can help you achieve that with five steps:

  1. Highlight all important paragraphs while reading.
  2. Dissect the content of a book into granular concepts.
  3. Map the relationships between these concepts.
  4. Group similar concepts together.
  5. Integrate these newly learned concepts with previously known concepts.

Here's a video of how I conduct this process. This video is not staged. It's a real example of how I develop an understanding of a book I read called Mindstorms: Children, Computers, and Powerful Ideas. You don't need to watch it fully (it's 4 hours long!), but quickly skimming through different parts of the video will give you a clear sense of how I extract core concepts from the book and develop my understanding of it. If you want to explore and play with the whiteboard I created in the video, here’s the link to it.

[2024/01/03 Update] I have created a tutorial showing and explaning how I conducted the five-step process in action using Heptabase. I highly recommend you check it out!

Step 1: Highlight all important paragraphs while reading.

The first step in the process is to highlight important paragraphs from the book you have read and organize them by chapter.

Depending on the reader you use, the highlighting process may vary. If you are using a desktop reader, you can simply copy and paste the text into a Heptabase card. If you are using a Kindle or iPad reader, you can export all the highlights of the book into a Markdown file and then import it into Heptabase as a card. If you are reading a physical book, you can take notes in Heptabase whenever you finish reading a chapter.

It doesn't matter which tool you use to create highlights, as long as your output is one large card for the book that includes your highlights from different chapters.

Step 2: Dissect the content of a book into granular concepts.

Once you have the book card ready, you can create a whiteboard and add that card through the import panel of the card library. In my example, I created a sub-whiteboard named Mindstorm under the parent whiteboard Reading Notes, and dragged my Mindstorms book card onto this sub-whiteboard.

Once I have the card ready, I open it on the right split panel so I can better see its content. I usually start by skimming through all the content to identify key concepts. When I decide that there's a concept I want to extract, I select the related blocks and drag them out onto the whiteboard to create a new concept card.

Simply creating a concept card is not enough. To ensure clarity, I always describe the concept in one sentence and use that sentence as the title of the concept card. Then, I reorganize the content of the card into a structure that makes sense to me, and perhaps even drag other related blocks from the original book card into this concept card.

Step 3: Map the relationships between these concepts.

As I extract more concept cards from the original book card, I gradually add connections between them, or place similar concept cards next to each other. If I notice two concept cards are about the same idea, I merge them into one. At other times, I might break down a large concept card into several smaller ones to ensure granularity.

Step 4: Group similar concepts together.

After extracting all the content from the original book cards into concept cards, I close the split panel and start working on mapping and grouping relationships. Often, I find multiple concept cards related to the same sub-topic. In such cases, I group those concept cards into a section and add a name to that section. Naming the section should be done as carefully as naming a concept card because these are the things that you recall first when you revisit this whiteboard in the future.

Completing steps two through four typically takes anywhere from an hour to a full day, depending on the length and depth of the book. Once you've produced the final whiteboard, the knowledge acquisition phase is complete. In this process, what's truly valuable is not the final whiteboard you produce, but the thought process you invest while establishing the knowledge structure and titling each concept card and section during steps two through four. Deep understanding and insights often come from the process of deconstructing, reassembling, and describing knowledge in your own words. Only after going through this process does the knowledge truly become your own.

After completing the final whiteboard layout (including all arrows and sections), I will open the original book card in the right split panel of the whiteboard and re-paste the links to all the concept cards and sections back onto this book card. In the figure below, you can see that each link in this book card displays the title of a concept card. This is why, in the second step, I summarize each concept card in one sentence and use that sentence as its title. Only by doing so can I see the core concepts of this book without having to open the links of these concept cards when reading the book card.

For example, if I name a concept card's title as "Engineers’ subculture," looking back a few months later, it would be hard for me to recall what this card is specifically about. But if I name the title of this concept card as "Engineers who find BASIC easy to learn formed a subculture that is influencing the world of education to favor students who are most like that subculture," even without reading the content, I can recall the core concept in the future just by its title.

Step 5: Integrate these newly learned concepts with previously known concepts

Up to now, I have gained a deep understanding of the core concepts of the book Mindstorms by breaking down and connect its core concepts on a whiteboard. But just understanding this book is not enough; I also want to truly integrate all the knowledge I have learned in the past, present, and future. In other words, I want to integrate the concept cards of this book with the concept cards I wrote for other books and lectures in the past.

Before doing this, I want to share an important learning mindset: You can only gain a deep understanding of the topics you care about through visual note-taking when you atomize your notes.

Many people, when first using a visual note-taking app like Heptabase, continue to use the old way of note-taking and write one note for each book or lecture, resulting in very long notes that contain many concepts. When your notes are in this format, it's hard to gain value from visualization.

For instance, in the figure below, there are two book cards, each with a lot of content. Although the content of these two books is related, connecting these two book cards on a whiteboard does not provide any new value, as it is almost the same as putting them in the same folder.

The purpose of visualizing notes is to gain a deep understanding of what you've learned. If all of your notes are very long and you don't break down the knowledge into smaller parts, the understanding you can gain from visualization will be very limited. Real deep understanding doesn't come from the "relationship between two books" but from the "relationship between all the concepts in these two books." What you want to connect are not book cards, but individual concept cards that you extract from these books using the previous four steps.

For instance, I've recently been researching how to design a computer-driven dynamic medium, and both books Mindstorms and The Early History of Smalltalk are highly relevant to this research topic. To better conduct this research, I created a whiteboard called Dynamic Medium and reused concept cards related to Dynamic Medium from these two books, organizing them using a mindmap to establish a unique understanding framework.

It's because I extracted and atomized the most important knowledge and ideas from these two books into reusable concept cards in the past when I read them, so that I can now easily apply my previous learning to new research topics. My past knowledge no longer sits uselessly in folders, but instead becomes the foundation for my newest research work!

Note: Atomic note-taking does not mean you cannot have long notes. It means that each concept card should only contain one concept and be supported by its content. If the content is long but all of it can be used to support the concept in the title, then the card is still an atomic concept card.

How to Learn and Research

After sharing how I implement the learning methodology in Heptabase, I would like to summarize the core ideas that underlie this methodology:

  • I believe that the essence of learning and research is to break down and extract the important concepts from books, literature, lectures, and experiences. Then, one should connect, understand, and internalize these concepts in one's own way to build a deep understanding of what is known and unknown to humans.
  • I believe all work plans and research papers are simply products of transforming this deep understanding into executable and communicable forms.

Under these core ideas, the processes of learning, research, planning, and output can be fully presented in the following diagram:

In this figure, on the left is source, which are the "literature cards" you wrote down while reading or attending lectures.

During my learning and research process, I extract useful concepts from these literature cards to create atomized "concept cards." Each concept card describes the concept in one sentence in the title and cites content from one or multiple pieces of literature to support this sentence. Every citation deepens my understanding and reflection of this concept.

As I learn and research, the original content of the literature cards will gradually be replaced by links to many concept cards. As I extract these concept cards from the literature cards step by step, I need to connect and fit them into a structure that makes sense to me. I can only truly understand and internalize a topic when I find such a structure for it.

In the future, whether I'm writing academic papers, work plans, or online articles, my process will involve linearly reassembling these concept cards into an "output card," which is an article meant for others to read. As I absorb and break down more and more "sources" during research, my "understanding" in the middle will deepen continuously, and the "output" on the right will naturally be of higher quality.

Closing thoughts

Although the topic of this article is about my method for acquiring, retaining, and applying knowledge, I do want to stress the importance of choosing the right tool to do so, because the design of a tool can radically change the way we subconsciously approach learning and form good and bad habits for it.

Seymour Papert, one of the pioneers of artificial intelligence and the constructionist movement in education, discusses his thoughts on this in Mindstorms:

For me, writing means making a rough draft and refining it over a considerable period of time. My image of myself as a writer includes the expectation of an “unacceptable” first draft that will develop with successive editing into presentable form. But I would not be able to afford this image if I were a third grader. The physical act of writing would be slow and laborious. I would have no secretary. For most children rewriting a text is so laborious that the first draft is the final copy, and the skill of rereading with a critical eye is never acquired. This changes dramatically when children have access to computers capable of manipulating text. The first draft is composed at the keyboard. Corrections are made easily. The current copy is always neat and tidy. I have seen a child move from total rejection of writing to an intense involvement (accompanied by rapid improvement of quality) within a few weeks of beginning to write with a computer. Even more dramatic changes are seen when the child has physical handicaps that make writing by hand more than usually difficult or even impossible.

Word processors can make a child’s experience of writing more like that of a real writer. But this can be undermined if the adults surrounding that child fail to appreciate what it is like to be a writer. For example, it is only too easy to imagine adults, including teachers, expressing the view that editing and re-editing a text is a waste of time (“Why don’t you get on to something new?” or “You aren’t making it any better, why don’t you fix your spelling?”).

As with writing, so with music-making, games of skill, complex graphics, whatever: The computer is not a culture unto itself, but it can serve to advance very different cultural and philosophical outlooks.

When building Heptabase, we aimed to design an environment that empowers you to externalize the process of identifying, dissecting, connecting, and grouping the concepts you have learned. That's why we built features such as the ability to dissect cards, move blocks across cards, build card relationships on a whiteboard, and reuse cards across multiple whiteboards. Together, these features form an environment that leverages the human capability of visual comprehension and visual memory with the computer's capability of data persistence and retrieval. With continued use of the tool to create understanding of your learning, you will start subconsciously adopting the habit of using the learning method described in this article. This is what ultimately matters—not just helping you take notes, but helping you become better at learning.







在這篇文章中,我會分享我自身學習的實際案例,展示我設計的一個用 Heptabase 來有效獲取、留存和應用知識的方法。這個方法雖然不是學習的唯一方法,但是一個我驗證過極為有效的方法,而且我相信大部分的人都可以很快地學會將它應用在自己的學習中。



  1. 將閱讀的過程中看到的所有重要段落記錄下來
  2. 將紀錄下來的重要段落拆解成顆粒度更小的概念
  3. 畫出概念之間的關聯性
  4. 將相似的概念群組起來
  5. 將這些新學到的概念與過去所學的已知概念整合

下面這支影片是我實踐前四個步驟的過程。這支影片是一個真實案例的錄影,完全沒有經過事先的規劃。在這支影片中,我示範了我怎麼拆解 Mindstorms: Children, Computers, and Powerful Ideas 這本書的概念、獲得深度的理解。你不需要把影片看完(因為它長達四小時),但快速的看過影片中的不同段落會讓你更清楚我從書本提取想法和建立理解的方式。如果你想玩玩看我在影片裡建立的白板,可以點擊這個連結

[2024/01/03 更新] 我錄了一支教學影片示範並講解了我使用 Heptabase 執行這五個步驟的方式,大力推薦你看一下。


在我的方法論的第一步中,我們會需要把在閱讀過程中看到的重要段落記錄下來並且按照章節整理。這個過程的實作方式可能會根據你使用的工具而有所不同。如果你用的是電腦的閱讀器,你可以直接將文字從電子書複製貼上到一張 Heptabase 的卡片裡頭。如果你使用的是 Kindle 或 iPad 閱讀器,你可以把所有 Highlight 匯出成 Markdown 檔案再匯入到 Heptabase 裡頭。如果你讀的是實體書,你可以在每一個章節讀完時做一次筆記。



當你將讀書筆記整理到一張書籍卡片以後,你可以創建一個白板,並透過 Heptabase 右上角的 Import Panel 將書籍卡片從 Card Library 匯進這個白板裡頭。舉例來說,我在 Reading Notes 這個母白板下創建了 Mindstorm 這個子白板,並且將 Mindstorms 這本書的卡片筆記放到了這個子白板上。






在將所有書籍卡片的內容都轉成概念卡片後,我會關掉右側欄,開始專心把這些概念卡片之間的關聯性和群組關係建立起來。當我發現有多張概念卡片都跟某個子題有關時,我會將它們用 Section 包起來,並給這個 Section 一個名字。我在為 Section 取名時會跟為概念卡片下標題時一樣謹慎,因為這些名稱將會是未來回顧這個白板時第一眼看到的東西。

完整走完第二步到第四步通常會花上一小時到一天的時間,這個時間取決於這本書的長度和深度。當你產出最終的白板後,吸收知識的階段就結束了。在這個過程中,真正有價值的並不是你最終產出的這個白板,而是你在執行第二步到第四步的過程中建立知識架構、給每張概念卡片和 Section 下標題所投入的思考過程。深度的理解和洞察往往源自於將知識分解、重組、用自己的話語描述的這個過程。只有在走過這個過程後,這些知識才會真正變成你的知識

在完成最終的白板排版(包含所有的箭頭和 Section)後,我會把原本的書籍卡片開到白板的右側欄,並將所有概念卡片和 Section 的連結重新貼回這張書籍卡片上。從下圖可以發現,這張書籍卡片中的每個連結顯示的內容都是某張概念卡片的標題。這也是為什麼在第二步的時候,我會將每張概念卡片用一句話總結,並用這句話當作它的標題。唯有這麼做,我才可以在看書籍卡片時,不用把這些概念卡片的連結點開就知道它的重點是什麼。

舉例來說,如果我把一張概念卡片的標題命名為「工程師的次文化」(Engineers’ subculture),過了幾個月後回頭看,我很難想起這張卡片具體在講什麼。但如果我把這張概念卡片的標題命名為「覺得 BASIC 語言很簡單的工程師形成了一種次文化,這種次文化影響著教育界,導致教師們偏愛那些喜歡這種次文化的學生。」(Engineers who find BASIC easy to learn formed a subculture that is influencing the world of education to favor students who are most like that subculture.),那我未來回頭看時,就算不看內文也能回想起它的核心概念。


截至目前為止,我已經透過在白板上拆解並重組 Mindstorms 這本書的核心概念,對這本書的知識獲得了深度理解。但是光是理解這本書還不夠,我還想要真正做到將我在過去、現在和未來所學的知識全部整合起來。換句話說,我想要將這本書的概念卡片與我以前為其他書和課程所寫的概念卡片整合在一起。


很多人在第一次使用 Heptabase 這種視覺化筆記軟體時,會沿用舊的思維為每一本書或每一堂課寫一個筆記,這些筆記的內容都非常長、包含非常多要點。當你的筆記是這種型態時,你就很難從視覺化中獲得價值。



舉例來說,我最近在研究如何設計一種以計算機驅動的動態媒介,而 MindstormsThe Early History of Smalltalk 這二本書的內容都與這個研究主題有高度相關。為了更好地做這個研究,我創建了一個叫「動態媒介」的白板,並將這二本書中與「動態媒介」有關的概念卡片復用進來,使用心智圖的功能去組織它們,建立一個我自己獨一無二的理解架構。




前面講了學習方法論在 Heptabase 的實踐,現在我想來總結一下這套方法論底層的核心思想:

  • 我認為學習和研究的本質是將不同書籍、文獻、課程、經驗中學到的重要知識和想法拆解出來,用自己的方式去關聯、理解、內化,進而對人類已知和未知的事物建立深度理解。
  • 我認為工作時寫的計畫和研究時寫的論文都是在將這些深度理解轉化成可以被執行和傳播的形式。








Seymour Papert 是人工智慧和教育建構主義運動的先驅之一,他在 Mindstorms 這本書中探討了工具如何影響思考的想法:




在打造 Heptabase 時,我們的目標是設計讓你可以將大腦學習知識的過程外部化的環境,讓你可以用眼睛和手去對概念進行識別、拆解、連接和分組。這就是為什麼我們開發拆解卡片、在卡片之間移動區塊、在白板上建立卡片關係、在多個白板之間重複使用卡片等功能。這些功能共同構成了一個環境,讓你能更好地運用人類的視覺理解和視覺記憶能力以及電腦的資料持久性和檢索能力來學習。當你愈常使用這個工具來對你的所學建立架構,你就會下意識地將這篇文章中描述的學習方法變成自己的習慣,而這才是最重要的 — 工具不只能讓你記筆記,更讓你成為一個擅長學習的人。