It’s hard to believe that I first accessed ChatGPT towards the end of 2022 and, if you read my original post, I didn’t think the technology was quite there at the time. But it nevertheless impressed me enough to know it was something I had to keep an eye on.
I’m thankful that I realised that early, as by 2024 I was using AI to help me edit my blog posts and by the summer of 2025 I had moved on from Grammarly entirely.
During the autumn and winter of 2025, I started experimenting with both Gemini CLI and Claude (desktop and Code) to see if these models could interact directly with my vault.
The short answer is they can. You can expect future posts on how I’m using them and the experiments I’ve been running.
As I continued to use AI, I could see its potential as a personal assistant and realised that for it to work at its full potential it needed access to both my PKM and my journals to give it full context on how it could help me. But there was a secondary concern: how do I keep my cognitive skills sharp?
Research seems to suggest that these concerns are justifiable, but it also suggests that if you use AI intentionally you can use it as a tool to improve your cognitive skills. This is what I’m hoping to explore in this blog post.
As this timeline shows, AI is developing very quickly and we need to start thinking about building an AI literacy to allow us to use our PKMs more effectively.
The Double-Edged Sword: Understanding Cognitive Offloading
Cognitive offloading occurs whenever you delegate a mental task to an external tool. Examples of this include:
- Jotting down a reminder to do something later
- Using a map application to navigate
- Using Google search for memory retrieval
It isn’t a recent phenomenon. It is something that has concerned people since the invention of writing — Socrates himself expressed concerns about the written word. When I was a child growing up it was television. Then it was the Internet, and our current concern is with AI.
In itself, cognitive offloading isn’t necessarily a bad thing, even though it does have a cost. Jotting down too many reminders can weaken your short-term memory. Using a map application too much reduces your spatial awareness.
But it also has a potential benefit: it frees up limited cognitive bandwidth that can be used for other cognitive tasks. Or not.
What are you replacing it with?
If you use this freed space for more demanding cognitive work, then your cognitive abilities will not decline as such, as you are still exercising your thinking.
However, you still need to be aware that whatever you decide to offload will weaken over time, as you are no longer exercising that mental capability.
Establishing the Boundaries: Your Red Lines
I would advise you to create your own set of red lines — a list of cognitive activities you will continue to do yourself. If your PKM is built on the Zettelkasten method, I think mine would make a good starting point:
- I will write my own literature notes
- I will write my own permanent notes
- I will add my own backlinks to my permanent notes
Those three red lines are where the important friction in my Zettelkasten happens. It’s that friction which helps me turn external ideas into my own personal knowledge.
The AI PKM Literacy Framework
That is why I have been working on an AI literacy framework for how you can use AI and a PKM together to improve your thinking. As someone who has kept a Zettelkasten for almost five years, it seems the most obvious source for a personal AI assistant.
Your Zettelkasten is a schema of your thoughts and ideas. It acts as a map, giving the AI context on who you are and what you think about. But how would this personal AI assistant impact our cognitive thinking?
It was that question which led me to this framework.
Due to research I have recently read on the expertise paradox, I would recommend that Levels 1 and 2 are only used by those with both AI and PKM experience. You can learn more in my post How will you know your PKM is ready for AI.
Let us take a look at each level in turn.
Level 0: Strategic Delegation
This level of the framework is about using AI to help with the low-value administrative tasks around keeping a Zettelkasten, with the aim of freeing up cognitive energy for thinking. Personal examples include identifying possible tags for a note.
You can learn more about this level in my post Introduction to my AI Knowledge Framework Level 0.
Level 1: Critical Refinement
This level is about using AI for summarisation as and when needed to help you understand complex or long source material. I myself don’t use it with every fleeting note — only those which are long or complex, after I have originally highlighted them.
You can learn more about Level 1 in my post AI Knowledge Frameworks Level 1-2.
Level 2: Socratic Interrogation
This is the most interesting level, where you can use AI as a thinking partner. This has two benefits: firstly, you have to process your thoughts into our abstract written language, and secondly, the AI might mention something you would never have thought of.
The important thing is that this is the start of the journey, never the end.
You can learn more about Level 2 in my post AI Knowledge Frameworks Level 1-2.
A Living Framework: Adaptability and Evolution
This is a living framework which will change over the coming months and years as both the capabilities of models and the results of research on the impact of AI and technology in general are carried out and published.
That is in part why I will be carrying out a quarterly review on how I use AI, to monitor how trying various techniques and models impacts my experience.
But the review has another aim: to ensure that I’m not becoming too dependent on the technology — a possible indication of cognitive offloading which could reduce the effectiveness of my own thinking.
To keep up to date with our latest content on this framework and related content around PKM, AI, and other technology-related posts, join our monthly newsletter.
Further Reading
- ChatGPT my first thoughts: Gives an idea of how quickly the technology has developed
- Gemini CLI and Obsidian Bases: A Showcase of LLM Strengths and Weaknesses in 2025: Gives an insight into how I was experimenting with Gemini CLI
- How will you know your PKM is ready for AI
- Introduction to my AI Knowledge Framework Level 0
- AI Knowledge Frameworks Level 1-2