Given our limited biological memory capacity, we have utilized tools to augment ourselves since the ancient times. Today, our computers, phones, and the Internet are now all considered extensions to our brains, where we can tap into almost the entirety of the collective human knowledge. All except our own personal knowledge that is, things we’ve learned in the past, people we’ve met, experiences we’ve had. I’ve been interested in ways to better manage personal knowledge for the last few years, and although there is a plethora of tools already (Org Mode, Roam, TiddlyWiki, just to list a few of the better ones), I don’t think anything that exists today quite do the job well.

First, is our personal knowledge even worth managing? The answer is of course subjective, and when it comes to how well one thinks the personal knowledge needs to be managed, we all fall on a gradient. Some are okay with a simple pen and paper notebook that has all the recent knowledge in it, while others diligently write, organize, and archive everything they’ve ever encountered. I think that our personal knowledge, if managed well, is extremely valuable in ways that we can’t even imagine today. In a way, it’s similar to asking whether it’s worth it to collect behavior data on the Internet: the answer changed dramatically after we’ve built the tools to extract value out of it. And for personal knowledge, we are still in the infancy of tool building.

Mindlessly hoarding notes is not an effective way to manage personal knowledge. I’ve done this myself in the past, and it’s very easy to get into a mode where I write down literally everything, then marvel at the huge collection of text that I’ve amassed with a sense of accomplishment. But information density is low, and more importantly, information retrieval becomes more difficult the more notes are added.

So we need to organize our knowledge. We can categorize personal knowledge by the different contexts: school/work related, family/personal life related, as well as the different types: knowledge learned, experiences gained (in diaries). We can record the relevant information and place them in the correct bucket for easier retrieval. We can have multiple layers of sub-categories (e.g. Learnings -> Science -> Physics -> Quantum Mechanics) for more granular organization, forming a complex tree of knowledge. This adds some overhead to simple note taking, but makes retrieval much easier, provided we know the context and type of information we are looking for.

However, knowledge isn’t really structured like a tree of categories. Many pieces of knowledge are related and interconnected, despite technically belonging in different contexts or categories. Tools like Roam and TiddlyWiki are created based on this realization, that knowledge is more like a networked graph. Each node of the graph represents a piece of knowledge, and it can be connected by edges to any other node in the graph regardless of its location. This introduces another dimension for which knowledge can be organized. If we imagined the tree of categories as one dimension, then the edges in the knowledge network form an entirely independent dimension. For example, a piece of knowledge about the lack of reproducibility in a psychology experiment from my spare time readings can be linked to the inconclusive results of a UX A/B test that I got at work, as they may have the same fundamental root cause.

The problem with this approach, as I’ve found, is the massive increased overhead to maintain such a comprehensive 3D knowledge tree-graph. Even if one takes the time to organize everything flawlessly once, it will degrade in quality over time unless the same level of effort is put in continuously to upkeep it. It’s definitely doable for individual projects, such as using Roam to organize knowledge this way on a piece of collaborative research work. But for the general ongoing collection of personal knowledge, this simply isn’t realistic for most people, myself included.

But if what we have is a collection of semi-structured network-connected text data, isn’t the problem already solved by Google, at least partially? After all, the Internet of hypertexts is exactly the same type of data, and Google does quite an impressive job at algorithmically organizing as well as finding everything. So maybe we can go back to simply hoarding notes, then run it all through the same intelligent indexing engine that Google has built for their web search, and query it the same way we search the Internet. Just as we can ask Google “who was the 42nd President of the United States” and get a direct answer right away, we can ask this personal knowledge Google “who was that tall guy I met at the airport last March” and get the answer, without having to do any tedious manual categorization and linking.

For this to work, we need two things: the recording of our personal knowledge, and the Google-like indexing and querying engine. The real Google only deals with the latter piece, as it relies on everyone else to create the content on the web, and it simply indexes and provides search over everything. But for our personal knowledge management system, we need the first piece as well, to record everything we need. Active note taking may be fine for some people, but not everyone wants to invest such high efforts. For something to become truly ubiquitous, it must have a low barrier of entry, and nothing is lower than a fully passive solution.

Suppose we had our phones in always-on recording mode, transcribing everything we hear and say. It will automatically know the contexts of interactions based on the location and who’s speaking (from voice detection). Better yet, in the near future we could all be wearing augmented reality glasses or contact lenses capable of recording everything we see as well. Biometric data, such as heart rates and mental states, can be collected as well from the sensors we wear. All this is aggregated and fed into a system similar to Google today, becoming fully indexed and queryable. It’ll be an extremely powerful personal knowledge management system that requires little to no overhead in organization. The true endgame here would be a full-fledged personal AI for each person, not just capable of answering questions asked, but also prompt conversations proactively.

For now, it seems that graph-based note taking apps are all we’ve got.