I had a conversation with Copilot about using Claude in a Personal knowledge management system. The question came from reading Sébastien Dubois’ article Agentic Knowledge Management: The Next Evolution of PKM where a series of automation activities are proposed.
There is a very strong line that I do not want AI to cross in The Quantum Garden Vault. Even with a case for AI’s automation to help me with knowledge management tasks AI must never create knowledge.
What does this mean? The knowledge in my vault is my knowledge, generated from my thinking and processes: not placed there by AI, copy/paste, attachments (PDF, epub) or ingested articles. It is important that I maintain Epistemic integrity and text represents my thinking and the knowledge I have created.
Any automation of “knowledge creation” by AI is not my knowledge and never can be. My knowledge comes from my owns words. It is where the learning happens.
I’m not yet sold on the idea of bringing AI to my vault as I only started thinking about it this afternoon. Here are a few possible uses I can think of:
- Automatic clean-up of notes missing information e.g.. books, (but what about Keep book list or not?). Outside of this, it never creates content
- Using AI as a second person who can look at what I wrote and challenge it for gaps, or Adversarial probing. I then have to be disciplined to fill the gaps myself and be aware that anything AI gives me is by its nature a summary, so I have to look deeper.
- AI can help me extract themes from what I’ve written as done with identifying some Initial interstitial journal observations.
