Safe Harbors for Deep Thought

PKM Tools are as Vast as the Ways We Work with Knowledge; this is Not Going to Change

One of the great challenges of PKM (personal knowledge management) systems is the vast variety of ways that people capture, hold, and access personal knowledge. It could be in Markdown format, stored in note-taking systems, stored in file systems, email, or even in spreadsheets.

Despite decades of attempts to homogenize personal knowledge management techniques, we’ve failed. Personal and corporate amnesia are common. We forget what we know. The diversity of PKM approaches continues to expand and LLM (large language models) is accelerating this expansion. I don’t believe humans will ever agree on a handful of ways to organize their digital work, let alone a single, unified approach to greater personal knowledge management.

It’s been said that

Email is Where Knowledge Goes to Die⁠

. I’m now getting the sense that knowledge dies a little everywhere it happens to live and partly because ...

There's no common interchange mechanism between all of these potential locations of personal knowledge.

Example: When I’m working in

Coda⁠

, I can’t easily inject an observation I made in an email message to a colleague, or a specific topic I explored in Tana.

Safe Harbors

Planet Earth is rife with commerce and where does it occur predominantly? Safe harbors - the calm waters that provide the ideal topographic marketplace where we can focus on transactions. Proximity of buyers, sellers, and their goods is tantamount to market mechanics. Knowledge transactions, which are represented by the data we choose to institutionalize for personal (or enterprise) benefit, also share this pattern. Largely, this is achieved through human recall, automation, or APIs.

Vastly, we assume that knowledge transactions are made possible through APIs. This works quite well to a point that mostly favors structured data. Humans, however, are required to increasingly establish loftier plateaus of knowledge with structured and unstructured information. LLMs, of course, play nicely in this work. NLP may provide the backdrop for deeper, if not quicker understanding.

But there's another type of transaction that we often underestimate in terms of importance.

Knowledge that moves from ourselves to ourselves.

Knowledge work is often an individual process that involves careful thought; synthesizing ideas; establishing connections; exploring hypotheses. These patterns are why we seek out tools like Obsidian, Mem, and Tana; each are safe harbors where we conduct personal knowledge transactions. They help us recall, link, organize, index, search, and distill ideas. But they are largely ineffective when they need to work together; the barriers and friction are significant. Interchange of known things is not possible from me to me when using different tools.

I find it a bit humorous that we’re so deeply focused on collaboration as the key to advanced thinking and knowledge when we’ve missed the most crucial PKM task - free and open interchange with ourselves in any application context.

The Interpersonal API

I have a long and storied history of hammering on PKM vendors for failing to build out their offerings based on their own RESTful APIs. While there are significant benefits to be gained by designing systems from the API up, this is not what’s important if the goal is to bridge the universe of PKM tools for our own internal deep work advantages.

RESTful APIs are not the answer. We need a methodology that is actually API-averse because APIs are a tax on knowledge transactions. They require lots of effort to design, build, and maintain. I believe LLM’s, embeddings, and vector databases are the raw materials required to design the knowledge adhesives we need.

The future interpersonal API is a vector space where everything you’ve captured — regardless of application used — is able to recall, blend, and capture new thoughts and ideas.

Next step - a prototype.

⁠

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.