My Path to Personal Knowledge Management (PKM)

I spent years trying to build the perfect system. New app, new hope. More platforms than I care to admit. Notebooks that lasted a few weeks. Systems that gave me that path to a perfect replacement for it all.

The hard part was always how people said I was so organized, and yet the system I live in – my own mind – kept breaking down. Here’s what I eventually figured out.

I have two corollaries about my memory:

  • I can’t remember anything without a point of reference
  • I can remember anything when I can draw it (find a mental model) or connect it to other dots (find the pattern)

This shapes everything. The problem was never the tools. It was that I kept building systems that worked against how I actually think. Here are 3 concrete takeaways from 20 years of these systems.

TL;DR

  1. Choose your why you’re getting organized to stay grounded
  2. Pick an intentional methodology of how to organize
  3. Only then choose a tool for the feeling it gives you
  4. Lastly, use AI for synthesis and reflection

Use it, tune it, but remember to use it more than tune it.

See Your Invisible Ontologies for

Developing specific categorical frameworks (ontologies) for the purpose of creating organizational patterns (structures).

Like most knowledge workers who care about their setup, I’ve cycled through my share of approaches as often as the tooling. (Dates are my usage, not the systems themselves.)

Getting Stuff Done (2015-2019) was the first one that stuck. Task capture, trusted system, weekly review. Useful for a while. But GTD’s entire orientation is toward the next task. It kept me efficient without growing me. I started to feel like I was feeding a machine.

Interstitial Journaling (2019-) was a genuinely good idea I still use – writing between tasks to capture the thread of thought that would otherwise disappear. Small but important. Not a full system.

Zettelkasten I read about extensively and decided wasn’t for me. The theory is beautiful. The practice never clicked with how I actually work.

Life OS with August Bradley (2021-2023) – I got deep into this one. Interconnected Notion databases, quarterly reviews, life pillars. Impressive to look at. But anyone who’s normalized a dataset would find the design offensive. The system was managing itself. I spent more time maintaining the structure than benefiting from it. (It also has a certain cult energy I should have clocked earlier.)

PARA (2023-2024) from Tiago Forte was the most practically useful. Projects, Areas, Resources, Archives – a sensible default for anyone who works in projects. “When in doubt, archive it” is genuinely good advice. But the Area/Resource distinction kept blurring, and eventually I realized: I was spending energy sorting things I wasn’t actually using. Sorting is not thinking.

Digital Garden <

Models for Task Management Specifically

Task management is its own subdomain of PKM and not something to confuse with knowledge management as a whole. It’s again not about choosing a system to be effective, it’s about finding one with categories that match how I naturally think.

The Many Methodologies of My PKM

  • Daily Flow for time as my default mental model (and the Obsidian plugins that make it possible)
    • Daily notes as a core interstitial journal (Period Notes)
    • Process for low friction capture (QuickAdd)
    • Custom views for feedback loops (Dataview)
      • Clarity of focus
      • Visibility progress
      • Interesting patterns
  • Creative Flow to remember I’m a maker, not only a consumer
    • Draw and write freely but backed by a system to find opportunities for reuse
    • Identify opportunities to share publicly without complexity and with assurances of privacy when I need them
    • Space to tinker with code contributions in minor ways to open source (thank you Obsidian community)
  • Knowledge Management helps by centralizing then connecting related information on engaging with a topic
    • Major change (Jan 2025): I’m shifting from a “gather and organize” to “recognize patterns and experiment”
  • Task Management, used minimally, tracks what I commit to and complete the most impactful work. Project management is a subcomponent to this larger theme.
  • Portability
    • Systematize preferences for reuse across environments (dotfile management)

Just as Important: Non-Goals

What once interested me but is no longer what I need:

  • Productivity optimization — I now work with my energy rather than rely on willpower
  • Digital hoarding – clipping everything I can to feel like I learned it

Choosing Tooling to Feel

The tool graveyard is extensive:

  • Omnifocus (2005-2007)
  • Evernote (2007-2018) ☠️🪦 < this hurt
  • Trello (2012-2021)
  • Asana (2013-2017)
  • Todoist (2017-2023)
  • Notion (2019-2023)
  • Reminders (2023-) < on and off, need a better option
  • Obsidian (2023-)

I have spent two decades now platform-hopping, but now I get why. It was invisible choices around the platform, not the software itself. I’ve mentioned the main ones: ontology, methodology, flows. But there’s also a feeling I noticed I’m chasing – it’s an aesthetic at the intersection of playful and prettiness.

Playful is personal. Mine feels like the inspiration to say “why not try” instead of “what’s the right way.” It’s a belief that everything can be a learning moment rather than a frustration. It’s safe and mischievous.

Prettiness is personal. Mine feels like flexible fonts, color schemes, and spacing to give me a different sense of space (so I can get into a different type of headspace). Mine is also full of bright colors and iconography that – for reasons I will never know fully – help me think.

Today I’m all about Obsidian. It works for me not because it’s the “best” app (though it’s incredible on so many levels), but because it matches how I feel about thinking: connections over categories, plain text that travels anywhere, plugins that extend rather than constrain. The graph view is less useful than advertised. The ability to link anything to anything is genuinely transformative.

The Right (and Wrong) Way to Add AI

Take this note with the appropriate level of context – thoughts on AI are sure to be out of date by the time I press publish. That acknowledged, I have found it incredibly powerful to find ways to connect my knowledge system with the powerful synthesis of ideas that today’s AI tooling can provide. Here’s how that evolved for me:

  • AI as a researcher: I had a thought, ask about related thinkers of related theories. Get a cool insight into how I think, save a few notes in Obsidian and then write what it means to me (to protect me doing the thinking, not outsourcing it).
  • AI as a synthesizer: I have thousands of notes in a known structure with a known methodology. I use a combo of an MCP connector and Claude Code in VS Code to look for interesting insights from themes to contradictions to unique thoughts I should consider writing as a blog post.
  • AI as a trusted agent (new & unstable): I have more ideas and tasks than time, so I’ve been playing with thoughtfully-scoped uses of agentic workflows from Claude Cowork to OpenClaw to OpenCode. I love the idea of agents helping me protect my attention by knowing what I value most (by telling it), giving feedback when my attention drifts, and then letting it help me stay organized along the way.

There isn’t a day that goes by without me plugging something into Claude alongside my PKM. I know that will be the norm going forward. But what I found quite quickly is the urge to “get organized” or “figure something out faster” has a dissatisfying conclusion for me. Claude following my lead and me guiding it to accomplish what I’m thinking is not the same thing as me growing through the process of thinking through something. I’ve read frequently about that false promise yet found myself falling for it too. It’s akin to how the Muller-Lyer illusion taught us that awareness of a bias will not prevent us from falling for it (there’s a fun Radiolab episode on the topic if you’re curious).

To say it simply, AI is life-changing for knowledge management. And it’s also a risk to your knowledge.

If you’re looking for a reflection tool and also using Obsidian, I can’t recommend this Claude skill by Marshall Kirkpatrick enough.

Finding My Current System

In early 2025, something shifted. The intersection of task management and knowledge management was a mess because I was treating them the same. Tasks are commitments. Knowledge is exploration. Mix them, and the infinite queue of things I could learn eventually crowds out the small number of things I’ve promised to do.

Old pattern:

  • Offload every wishlist idea into the system
  • Complete a fraction of them
  • Carry the rest as permanent backlog
  • Get overwhelmed
  • Add more anyway
  • Repeat until sad

New pattern:

Separate commitments from exploration. For real this time. Small things go in Todoist or Reminders. Knowledge management 95% of Obsidian now, but organized around ideas in motion, not topics to file away. I treat long-term commitments as knowledge to manage, not a task to complete.

For example: if I have a presentation to give on product operational models, I put that in Todoist with a due date. I also make a folder in my “Active” top-level folder where I build out my thinking, add research, and write up ways of learning by doing (more on that below). Both locations complement each other.

That’s evolved into this folder structure:

My Obsidian Folder Structure

  • _Capture is the default and filled with daily notes and quick thoughts
  • Active is project-oriented, but not a strict project management system
  • Archivist is all my past systems
  • Grafts are other people’s content I like to revisit
  • Knowledge is well integrated concepts I’m curating for myself
  • Reflections are new ways of analyzing year over year patterns
  • System is files, templates, and scripts

That structure shifts here and there as my focuses shift (e.g. Grafts used to be in Archivist), but are relatively stable. What makes that possible is the organizational theory underneath it: instead of organizing by topic (which just recreates my old filing sorting problem), I organize by state – borrowing an idea from enabling constraints:

  • Droplets – raw capture. Something caught my attention. No commitment to where it goes.
  • Emergent Patterns – I’m noticing a theme across multiple inputs. Still just watching, not filing.
  • Experimenting – I’m trying it. Living with it. Seeing what happens.
  • Ideal – A clear target state I’m working toward. What done looks like before I start.

This is emergence over hierarchy. Notes evolve through connections, only loosely depending on rigid folders. Structure serves discovery, not filing.

The other thing that changed: I stopped carrying debt. If something goes unfinished and I no longer care about it, I cancel it without shame. I even built a custom Templater function (fn Cancel Incomplete TODOs) to do this without ceremony. The debt is optional. I’d just never noticed I could put it down.

What This Is Actually For

I’m not optimizing for productivity – I tried that, and it made me unhappy in my system of thought, which made me avoid it. I’m not hoarding information – I tried that too, and forgot most of what I supposedly learned anyway. What I’m doing now is recognizing patterns and experimenting with what emerges.

My Obsidian vault exists because of those two corollaries. I can’t remember anything without a reference point. But once I can connect something to a pattern, or draw a mental model that holds it – I remember it. The goal isn’t the system. The goal is the retrieval. The thinking that happens when one note connects to another in a way I didn’t expect.

That’s it. Everything else is context.