On Note-Taking and the Myth of the “Second Brain”

For several years now, the idea of a “second brain” has become one of the most appealing promises in the world of productivity and knowledge management. The idea is simple: our memory is limited, information has become overwhelming, so we should move part of the mind’s work outside ourselves—to Obsidian, Notion, Roam, or any other software that can store notes, connect them, and later hand them back to us. If we build this system properly, supposedly we forget less, know more, and ultimately write more, think better, and produce more output in our field of expertise. I think the last part of this narrative is largely a myth.

Not in the sense that note-taking is useless, or that tools such as Obsidian do not work. On the contrary, outsourcing part of the memory burden is a wholly real and well-known cognitive behavior. Psychologists use the term cognitive offloading for it: using the external environment to reduce the cognitive demands of a task. Writing a shopping list, setting a reminder on a phone, or saving an academic reference are all examples of this. In their well-known review, Risko and Gilbert show that people regularly use the environment as an auxiliary memory. But this very research makes clear a boundary that often disappears in the “second brain” literature: reducing the memory burden is not the same as learning.

In several experiments, people who were allowed to entrust information to external memory performed better on the immediate task, but later remembered the information itself more poorly. Put simply, the external system had done its job well; what had not necessarily improved was the person’s memory. This finding is not, in itself, an argument against note-taking. If I have saved a phone number on my phone, there is no virtue in memorizing it. But if what I have outsourced is a concept I am supposed to think with, build an argument with, teach, or use to solve a problem, then the matter is different.

This is precisely the point at which the “second brain” metaphor becomes misleading. An information repository can be a good external memory, but external memory is still not “knowledge.” Knowledge is at my disposal when I can retrieve it, build relationships between it and other things I know, apply it in a new situation, generate questions with it, and defend it against criticism. Having five thousand notes on cognitive psychology does not necessarily make me someone who understands cognitive psychology better; just as having a large library alone does not make someone a thinker.

Epistemologically, too, this distinction predates our software. “Accessible information” is not the same as “knowing.” Even if we accept a very radical version of the theory of the “extended mind” and say that, under certain conditions, a notebook, phone, or software application can be part of our cognitive system, the question still remains: what does that system do with the information? Archiving a proposition does not increase our ability to reason with that proposition.

The main problem with many knowledge-management systems is that they reduce the entire cognitive cycle to the stage at which tools perform best: recording, categorizing, and retrieving. Folder structures improve, links multiply, the graph view fills up, and the number of daily notes rises. All of these are measurable, and that very measurability creates a sense of progress. But what is truly difficult is less visible: understanding, forgetting and recalling, explaining without looking at the text, discovering a contradiction between two ideas, changing one’s mind, writing, teaching, and testing or deepening one’s understanding of a concept.

Suppose two people read a book. The first takes a hundred detailed notes, creates several tags for each note, and links them to dozens of earlier notes. The second has fewer notes, but the next day closes the book and tries to explain to a friend exactly what the author has claimed, where the argument is weak, and what the book has changed in their prior understanding. If, a month later, we ask both of them to talk about the book without consulting their archives, it is by no means obvious that the first person—who may well have put more effort into collecting—has “known” more. Many findings in learning psychology actually support the opposite.

One of the most robust findings in this field is retrieval practice, or retrieval practice. It has repeatedly been shown that trying to pull knowledge out of memory is, in many circumstances, more effective for long-term retention than rereading the same material. Meta-analyses and classic studies have reported this effect across various domains. The interesting point is that learning occurs precisely where the task becomes a little harder. When the text is in front of us, we feel familiarity; when we close the text and have to ask ourselves, “What did I really understand?”, the gap between familiarity and knowing suddenly becomes discouragingly apparent—and perhaps, especially in the age of AI and the short-attention economy, we are not very interested in confronting our own mental sluggishness.

The same logic applies to generation as well. The literature on the generation effect has shown that in many situations, constructing or producing an answer can strengthen memory more than passive reception, although the magnitude and conditions of this effect depend on the type of task. Some studies even show that generative activities, such as making a sentence or creating a semantic connection, sometimes work better than simple retrieval. Therefore, the knowledge cycle cannot be reduced to “input → note.” The mind “learns” through transforming, retrieving, explaining, making mistakes, and correcting them.

This is where the relationship between “being taught” and “teaching” becomes important to me. We are accustomed to seeing these as two separate activities: first we learn something, and when we know enough, we teach it to someone else. But learning by teaching clearly shows that preparing to teach and the very act of explaining something to another person are important parts of the learning process. When we are going to teach something, we are forced to find its structure, see the gaps in our understanding, clarify the relationship among its components, and think about questions that we had not noticed at all while reading passively. The effect is not always the same under all conditions, but the research literature as a whole shows that teaching can be a serious learning mechanism for the teacher themselves.

For this reason, a truly knowledge-centered system must have “output” and “cycle” built into its architecture from the beginning. If I am going to learn something, it should be clear from the outset what I am going to do with it: explain it? write an article? use it in a project? discuss it with someone? solve a problem with it? criticize it? If none of these is involved, and an AI agent, instead of proposing such things, takes over the task and does it in our place, it is very easy for note-taking to turn into a pointless fad and trend.

Perhaps that is why some vaults, after a few years—if we have such perseverance—turn into warehouses rather than second brains. Thousands of objects exist in them, but the flow of intellectual life does not pass through this furniture and equipment. One can spend hours refining taxonomy, template, tag, and backlink, and at the end of the day feel that one has “worked” on one’s knowledge, while a large part of the time has been spent on “the architecture of a warehouse.” Someone with a large, well-organized collection of carpentry tools is not necessarily a better carpenter. Tools acquire meaning when we have something to build.

This error, of course, is not merely the product of the growth of the personal knowledge management industry. These tools have grown on a cultural ground that has made “receiving” far easier than “engaging.” The endless stream of articles, podcasts, videos, newsletters, highlights, and bookmarks constantly creates the feeling that there is always one more thing we must see before beginning the real work. In such an environment, attention itself becomes a scarce resource. Research on media multitasking and phone notifications shows that repeatedly shifting attention, and even receiving a notification without responding to it, can impair performance on tasks that require attention.

Therefore, perhaps our diagnosis of the problem has been somewhat wrong from the outset. Modern humans certainly face an enormous volume of information, and their working memory is limited; denying these limitations is inconsistent with cognitive evidence. But the main difficulty cannot simply be called “managing the volume of knowledge.” What has become abundant is information, not necessarily knowledge. And what has become scarce is not merely storage space; it is attention, active retrieval, enough time to combine ideas, and situations in which we are compelled to put what we know to use. From this perspective, the promise of a “second brain” is inverted. The second brain is not supposed to fill in for the first brain. A good external system should make the first brain work.

I still take notes and will probably continue to use tools such as Obsidian when necessary. For me, the issue is not note-taking itself. The issue is when a note re-enters the process of thought. Have I made a question out of it? Have I been able to reconstruct its idea without looking? Have I explained it to someone? Have I written something that could not have existed without that note? Has some new knowledge led me to revisit one of my previous notes? From this perspective, my argument is that today, given the issues raised about attention and social media, we should post notes rather than take notes; by this I mean that we should put a note on a wall for others to see. Stockpiling knowledge makes it far less likely to be seen again. Therefore, even in note-taking, if the answer to the questions above is negative, note-taking has none of the benefits claimed for it except for historiography and documentation. In the end, it creates a well-made archive, which is not a bad thing, but it should not be confused with knowing.

After years of trying to end my personal chaos in using tools for writing, publishing, knowledge management, and education, I have finally reached a result that I have never felt closer to in any previous project. I started the LeraLink project under another name—iLiBRi—in 2023. I had begun the effort to build or design a knowledge creation and management system a few months prior, back when Perplexity, OpenAI, and its new product ChatGPT did not yet have a wide introduction in the tech market. Over these few years, I participated in almost all communities of such tools and solutions, and was a primary user of all the new tools in this field: Zotero, Zenodo, PKP, Notion, all kinds of LMSs, as well as Obsidian and similar tools. I also employed various project management systems and open academic or non-academic publishing platforms, writing and SEO tools, and website designs that could facilitate knowledge creation and management. I explored all sorts of community features, PKM (Personal Knowledge Management), and project management tools, such as Basecamp with the Shape Up methodology, which I think is one of the best. Along this path, I also developed a great interest in Perplexity and how it works. Gradually, I became familiar with what are called AI-Native tools—where AI (or LLMs) was not merely a complementary, enhancing, or decorative part of those systems. These included mini-course generators and various simple and complex content creation tools, up until recently when I also became familiar with writing tools for fiction and non-fiction novels, especially those powered by AI that fundamentally emerged from the intersection of writing experience and AI technology. This might offer a small, albeit incomplete, perspective on the reality that LeraLink is born out of a multi-layered, vast, and multi-year technical, intellectual, experiential, and scientific journey, arising from numerous failed projects alongside scattered but successful small experiments. I hope you are interested in getting to know it. My ultimate decision in recent months was that in the process of building this federated knowledge creation and management system—which will gradually incorporate all key steps of the knowledge cycle and utilizes AI in a different way—I should be its very first user. If a tool is not useful enough for its creator, it certainly will not be for others[1].

اگر علاقه‌مندید، با عضویت در پلتفرم من در جریان توسعه و تغییرات پروژه بمانید. البته این هشدار را به شما می‌دهم که قطعات من تنها به این پروژه بسنده نمی‌کنند و در حوزه‌های گوناگونی کار می‌کنم.