🗃️ Zettelkasten Hub — PUMA Research Knowledge Base
Zettelkasten principle (Niklas Luhmann, 1927–1998): knowledge grows through atomic, permanently linked notes. Each note = one idea. Notes never move. Links create emergent insight.
This hub follows the Obsidian Zettelkasten Starter Kit pattern by Edmund Gröpl, adapted for the PUMA Project research context.
🗂️ Zettelkasten Note Types (adapted from groepl)
| Type | Prefix | Location | Purpose |
|---|---|---|---|
| Fleeting Note | FL- | 10 - Inbox/Fleeting-Notes/ | Raw capture, process within 48h |
| Literature Note | LN- | 20 - Literature/ | External source processed with Keshav |
| Permanent Note | PN- | 30 - Permanent/31–33/ | Your synthesised atomic idea |
| Structure Note | ST- | 30 - Permanent/30 Zettelkasten-Hub/ | Index / hub for a cluster of PN |
| Source Note | SRC- | 30 - Permanent/36 Sources/ | Full source metadata (author, year, context) |
| Person Note | PER- | 30 - Permanent/37 Persons/ | Key thinkers (Keshav, Flyvbjerg, Luhmann…) |
| MOC | MOC- | 80 - MOC/ | Navigation map across note clusters |
📋 PUMA Permanent Notes Index
31 Concepts — Core Ideas
TABLE file.ctime AS "Created", tags AS "Tags"
FROM "30 - Permanent/31 Concepts"
WHERE type = "permanent"
SORT file.ctime DESC32 Methods — Research & Statistical
TABLE file.ctime AS "Created"
FROM "30 - Permanent/32 Methods"
WHERE type = "permanent"
SORT file.name ASC33 Frameworks — Methodologies & Prompting
TABLE file.ctime AS "Created"
FROM "30 - Permanent/33 Frameworks"
WHERE type = "permanent"
SORT file.name ASC🔗 Zettelkasten Rules (PUMA)
Core Rules — Memorise These
- One idea per note. If a note wants to say two things, split it.
- Declarative title. The title IS the claim. Example: ✅ “RAG retrieval improves precision when examples are structurally similar to the query” ❌ “RAG notes”
- Never move a permanent note. It lives in
30 - Permanent/forever. Projects archive; knowledge does not.- Always link. Every permanent note must link to: its source LN, at least one related PN, and one MOC.
- Write in your own words. No copy-paste from sources. If you can’t restate it, you don’t understand it yet.
- Distinguish claim from evidence. State the idea, then support it. Never bury the claim in the middle.
🌐 Knowledge Graph Status
TABLE length(file.inlinks) AS "Inlinks", length(file.outlinks) AS "Outlinks"
FROM "30 - Permanent"
WHERE type = "permanent"
SORT length(file.inlinks) DESC
LIMIT 15🔍 Orphan Notes
LIST
FROM "30 - Permanent"
WHERE length(file.inlinks) = 0 AND length(file.outlinks) = 0
SORT file.ctime ASC📚 Source Notes Index
TABLE author AS "Author", year AS "Year", type AS "Type"
FROM "30 - Permanent/36 Sources"
SORT year DESC👤 Person Notes Index
TABLE role AS "Role", field AS "Field"
FROM "30 - Permanent/37 Persons"
SORT file.name ASC📎 Quick Note Index (permanent notes)
31 Concepts
- PN-IssueTriage-StoryPoints — Issue triage + effort estimation (F1-macro, MAE)
- PN-CoT-FewShot-Prompting — Chain-of-Thought + Few-Shot theory
- PN-KeyConcepts-Agents-Reproducibility-RedTeam — LLM Agents, Reproducibility, Uniqueness Trap, Red Teaming, Agent OS
- PN-LLM-Local-vs-Cloud — Why local inference for benchmarks
- PN-RAG-Embeddings-VectorDB — RAG for Stage 4
- PN-ReAct-AgentPattern — Reasoning + acting agent loop
- PN-MultiAgent-ArchitecturePatterns — Specialisation boundaries
- PN-Agentic-Science-Paradigm — Agentic science paradigm
- PN-AI-Scientific-Knowledge-Generation — AI-driven discovery
- PN-PUMA-within-AgenticScience-Trajectory — PUMA’s theoretical position
- PN-ToolSelection-PUMA — Tool choice rationale
32 Methods
- PN-DSR-SLR-Methods — DSR paradigm + PRISMA SLR
- PN-Wilcoxon-FINER-Cornell-PRISMA — Statistical methods + FINER
33 Frameworks
- PN-SDD-Framework — Spec-Driven Development + BDD + BMAD
- PN-RCOIF-Framework — Structured prompting methodology
- PN-MIT-Student-Method — Research reading method
- PN-EGI-Framework — Guided domain exploration
- PN-AMI-DRCA-IIPR-Frameworks — Advanced prompting techniques
- PN-PARA-GTD-Zettelkasten — Vault organisation integration
Structure Notes
- ST-Prompting-Strategies — Prompting strategies cluster
- ST-Reproducibility-Cluster — Reproducibility cluster