I have tried starting both analog and digital Zettelkastens several times but never kept up with them because the maintenance always seemed overwhelming. The "LLM Knowledge Bases" Karpathy tweet made me wonder if an AI-assisted version might work, and it worked beyond my expectations, so I thought others might find it useful.
You dump your sources in the raw folder and ask your AI of choice to ingest/process them. This produces a linked wiki of cited summaries, topics, and concepts. Then you can query for other kinds of outputs based on your sources, like research summaries, answers to questions, gap analyses, etc. You can later add additional sources to raw to be processed and integrated into your library.
You dump your sources in the raw folder and ask your AI of choice to ingest/process them. This produces a linked wiki of cited summaries, topics, and concepts. Then you can query for other kinds of outputs based on your sources, like research summaries, answers to questions, gap analyses, etc. You can later add additional sources to raw to be processed and integrated into your library.
I made a mini sample library so that you can see the kinds of outputs it produces. Browse the wiki and outputs folders in starter-library: https://github.com/ulyssestenn/funes-example-ai/tree/main/st...
Would be interested in your feedback.