Exploring the Intersection of AI and PKMs

How do you define a “Personal Knowledge Management” system, and what is its primary purpose in your life?

Let me actually draw on some previous reflections I've made in my notes about PKM's, making this the most meta definition ever....

Personal Knowledge Management is a private collection of information designed to serve a lifetime of learning and growth, not just a single use case. In addition to storing information, it also provides a studio for exploring, testing, and refining ideas or connections before committing them.

PKM's also extend into the workplace to support knowledge workers who need to be responsible for their own growth and learning.

It's not about having the perfect tools – it's about having a reliable set of tools you can depend on, knowing you can always change them later.

It's not about having the perfect framework either -- it's about having a system that matches the way you think like glove.

Your PKM should remove mental burden not add to it.

Your PKM System is a crucible for ideas. your attention – as an asset that gets invested and produces a return, which in turn can be reinvested back into other ventures. This is how you can ensure your knowledge grows and compounds over time like a high-yield asset.

Recommended Reading:

• Building a Second Brain
• Duly Noted

How does introducing AI into that system strengthen that compounding effect, and is there any risk that it might flatten it by shortcutting too much of the friction that produces insight?

The current PKM tools still have two great failings that prevent them from truly meeting our needs as knowledge workers:

1. Discovery
2. Abstract Connection

All of our tools attempt to solve for these in some way, but they all fall short.

For example, Discovery, that is, finding the information you need, still relies heavily on you the architect-ing your PKM's organization in a way that allows you to easily find content. This may be by adopting a specific system like LATCH or PARA, or it might be through the use of tags, or you may simply rely on the powerful tools of keyword search. But all of these options are a type of mental burden on the user, both at the time of content creation (deciding how to organize) and at retrieval (deciding which method to use).

Finding content in your knowledge vault can sometimes feel like playing 20 questions with yourself.

And when it comes to abstract connection, that burden still relies heavily on the user. You must remember that something exists that is indirectly related to what you are currently working on. Keyword search only goes so far, and in general you must resort to randomly reading over lots of content to see if it sparks any ideas or connections. This is fine for organic inspiration, but inefficient when trying to do intentional research.

For both of these problems, AI is a perfectly suited. AI is flexible enough to work across both structured and unstructured content and it can pull an obscene amount of information into its context window. This can alleviate the Discovery burden as organization becomes less critical.

AI can also attempt to form inferences, meaning it can be used to help brainstorm loosely connected concepts. Ideally, our PKM tools are serving as a kind of brainstorming partner, floating indirectly related content up to us to see if we think there should be a connection there.

However....

The fact that AI shortcuts much of the friction that honed our thinking power is a real concern. In fact, I'd like to write a dedicated blog post on this topic as I think awareness of this pitfall is critical in avoiding it. Remaining a power user of AI years down the road will hinge on your ability to keep your mind sharp and not let skills atrophy as AI fills more gaps in your workflow.

Reliance on AI to think for you is a real concern and something that needs to be thought about responsibly when applying AI to your PKM workflows. I think a good strategy here is to start with using AI for Discovery and also for streamlining as many repetitive, non-thinking tasks as possible (like admin work and organization) to free up your time and brain power for the harder work.

Can you walk me through one specific moment where AI meaningfully accelerated your thinking or output inside your PKM — something that felt like a turning point or a “whoa” moment?

My ground breaking, aha moment came at work.

For context, my work PKM vault is a collection of not only every project I've worked on, but also serves as expansive documentation across all the features and systems I've learned about at our company over the past 8 years. Its highly personal, and by no means in a publishable state, but you can think of it as my own personal company Wikipedia.

Considering the general challenge of getting Engineers to write documentation and keep it updated, especially in a fast paced/high turn over environment, I've found a PKM system to be the original accelerator to my working capabilities. In my head I always think my title should be "Knower of Things" instead of "Staff Engineer".

Naturally, for a while now I had wanted to point an AI Agent at my vault so I could ask it questions to find answers instead of my current process of searching and or trying to remember where the answer may live. As mentioned earlier, make Discovery easier.

However, I hadn't quite found the right tool for this. So far the only option I had at work was a license to ChatGPT Chat. And while I could upload content to create custom GPT's, it required regularly, manually, uploading the latest version of that content. This was a non-starter, I'm updating my notes all day everyday.

THEN, our company started testing out Cursor. Being the blasphemous engineer that I am, the first thing I used Cursor for was not writing code, but instead querying my notes. I opened up my vault notes folder inside of Cursor and started asking it questions.

👼 cue angelic voices singing 👼

Just like that, I could say "bring me up to speed about feature X", and not only would AI find the definition of what that feature is and how it was built, it would also find the recent inflight project note about a planned change to that feature.

Ask it to summarize and dump that into a markdown file and now I've got something super convenient and polished to share with my coworkers as knowledge transfer.

Using Cursor with Obsidian

Configuration and Rules I’m using to power my Cursor + Obsidian use cases

Next TopicBailey Belvis

This calls back to something I said earlier about how one of the purposes of a PKM was that it was an investment in future you. I've felt that payback multiple times since starting the PKM journey, but never quite so much as realizing how it prepared me to be a power AI knowledge worker now at my job.

Of course, this still all hinges on me having acquired the knowledge into my vault in the first place. I've been thinking a lot about how to democratize this kind of knowledge curation powered by AI.

In what ways has AI helped you explore, test, or refine new ideas — not just retrieve past ones? Have you ever been surprised or challenged by what it surfaced?

I'm still relatively early in my AI journey so much of what I do is experiment still. I haven't always identified the best way to do something, and with how quickly the landscape is changing it can be hard to keep up.

With that said, most recently AI has helped me explore and refine ideas by being a sort of thought partner.

Actually, this interview right now is the perfect example of using it for that purpose. The idea to have an AI Agent interview me about my thoughts on this topic was partially inspired by this agentic coding strategy along with other things I've read that had mentioned a similar approach for refining an idea.

I've been surprised at the quality of questions AI has chosen to ask. They exceed my expectations. I've also been reminded of this type of thinking as a lost skill. During my school years this was an activity I had to do regularly for writing papers and formulating opinions. However, as a software engineer, formal (non-technical) long form writing has become very dusty for me. So having AI coach me through the thinking the process again has been incredibly helpful.

Even within this interview I've discovered so many thoughts I knew I had floating in my brain but had not know how or when to express. Sometimes when I think about an idea or opinion, I feel like my brain shatters into a thousand different directions, each with its own interesting ideas and connections to explore. This is frankly overwhelming and can cause a sort of expression paralysis for me which is also not at all helped by my perfectionist tendencies.

By having this interview, I've removed some of my personal mental burden. I don't need to worry about following every stray thought, or finding a way to organize this madness, I just need to answer the question.

If AI can help you think, organize, and write — then why bother with a PKM system at all?

Here's the thing, AI, by default and for the most part, works off of publicly available knowledge.

Everyone who uses it, is pulling from this same pool of information. You can use it as a thought partner with just this. But if you want to get to the next level and differentiate yourself from other AI users, then what you bring to the process is the most important thing.

A PKM, is personal, its something no one else has. Even if two people explore the same topics in their PKM's, the way they explore, connect, and even organize that information will be different. And all of this contributes to how the person reasons about that information. AND it can influence how AI will reason about that information if provided to an agent.

This is currently my favorite quote as it relates to PKM's, AI's, and creativity:

Do you think AI will lead to bloated vaults full of unrefined content? Will it weaken the signal-to-noise ratio that makes PKMs useful?

Gosh this question hits hard. This is exactly right and exactly something I'm still thinking about actively.

Personally, the majority of my "uncurated" content originates from content (articles/posts) I read online. It is not completely uncurated, without going into too much detail, the process is something like this:

1. Subscribe to RSS feeds and get a fire hose of articles
2. Sift down to just the stuff I'm interested in reading
3. Read that stuff
4. Then sift down to just what I highlight, annotate, or want to keep for future reference
5. Sync only #4 into my PKM

In the old world, after step 5, I'd try to write some kind of summary note. I refined and distilled takeaway. But this was time consuming and honestly, rarely happened.

Now with AI, I can have AI do that "post processing" for me. I've already observed that the summary generated by AI is not as good or as useful as if I had written it myself. And it certainly always lacks my personal perspective. BUT, I'm starting to view this as more of a waypoint.

Specifically for source material like this (often also referred to as Literature notes), I want to create waypoints to ease discovery in the future without having to re-read every article. And for this, AI is adequate.

Ghostreader Has Earned a Place in My PKM Workflow

Using Ghostreader custom prompts to enhance my synced Obsidian notes

Next TopicBailey Belvis

I've gotten away from your question a bit, but the point of that was to illustrate that its possible we already had noise in our vaults to begin with, and in my case, AI can actually help add extra signal to that noise.

Curation and cultivation are the value add provided by a PKM to AI. Therefore, any time you plan to introduce AI to your vault, it must always be guarded by that first premise. You should strive to protect the curation of the vault over the excitement of AI.

As AI continues to integrate into PKM tools — what do you see evolving in your workflow over the next few years? What excites or worries you?

First, I expect AI to become a first class feature within all PKM tools, i.e. no longer bouncing back and forth between Chat interfaces or Cursor, or using weird third party plugins to get AI integration.

I also expect AI to become more proactively integrated. Similar to how AI code completions will write entire blocks of code at a time inline, I expect something similar to be offered in PKM tools. BUT, I don't expect that inline suggestion to be written content (we need preserve the purpose of the tool which is to capture the author's brain and words not the AI's). Instead I'm envisioning inline suggestions of provoking questions, semi-related content, editorial help. Like an all knowing editor looking over your shoulder as your write. One who is personally familiar with your style and way of thinking.

Regarding what excites me, I'm excited about the barriers to entry that may be broken down by AI. What new ideas will be brought into the light by people who previously didn't have access to the resources to be heard or seen? How will power dynamics shift as AI is brought to the general population? Is this a new type of information age revolution? What will those who have power right now do in order to keep their power? Should we be doing anything now to avoid AI provider monopolies?

Those final questions lead into my greatest fears when it comes to AI in general.

The first I've hit on already in previous questions which is brain atrophy. What skills will we humans lose as we increasingly rely on AI? The very skills that make us such powerful AI users today, could be lost 5 years from now.

My second greatest concern is privacy and security. PKM's can be exceptionally personal. Many choose specific tools so they can avoid their personal data being uploaded to a server somewhere. AI is generally not locally hostable at this stage. Which means that in order to leverage it, you are trusting a 3rd party with your information.

Many will dismiss that I am one of those "tinfoil wearing hat lizzard peoples", but my personal data in the hands of 3rd parties is one of my greatest fears. Data leaks, targeted advertising, government targeting... all the things we feared about our social media accounts being used against us in the past, are magnified 100x fold with AI's potential knowledge about us.

What kind of person becomes most powerful with a PKM + AI workflow?

Currently, the person who has a lifestyle of curiosity, discovery, and introspection has the most to gain by becoming a power user of AI. They are best primed to use AI + their PKM in a differentiating way.

Some guide posts I'm currently trying to follow:

1. Curate and cultivate your PKM garden, providing it a steady stream of fresh content and perspectives that resonate with you
2. Enrich your PKM to accentuate your voice, view points, and thinking on topics
3. Learn as much as possible about AI tools, what is available, which ones excel at which tasks, become an expert in the tools you choose to use
4. Learn about the risks of AI, and be on the lookout for how to avoid skill atrophy
5. Experiment and be curious

I also recommend checking out this article. It's a bit preachy, and I do not agree with everything in it, but I do think it does a good job of expanding the thinking about humans must bring to the equation going forward.