This is a prototype I’m using daily. I built a Zettelkasten-inspired note system that uses vector embeddings to suggest connections between notes.
It combines a Zettelkasten-like system with semantic search via Pinecone. When you add a note, AI suggests which existing notes it relates to. You approve the connections, and it maintains bidirectional links automatically.
Unlike traditional Zettelkasten, this treats tasks and knowledge notes equally. My tasks aren’t separate from my learning – they’re where insights emerge.
Stack: Pinecone, multilingual-e5-large, Claude MCP, shell aliases for quick prompts (zadd, zsearch, zboard).
The magic is in the system prompt (included in repo) that tells Claude Code the ID conventions, connection discovery logic, and workflows.
Early prototype, but the approach feels promising (at least for me!). Looking for feedback from PKM enthusiasts and folks who’ve built similar tools!
Comments URL: https://news.ycombinator.com/item?id=45459907
Points: 1
# Comments: 0
Source: github.com
