Knowledge Base

If the user doesn't specify, default to Introduction. Ask only if genuinely ambiguous.

2. Research

Use WebSearch to gather comprehensive information:

WebSearch: "[concept] definition explanation"
WebSearch: "[concept] examples use cases"
WebSearch: "[concept] advantages disadvantages limitations"
WebSearch: "[concept] vs [related concept]" (for counterparts)
WebSearch: "[concept] history origin year introduced paper"
WebSearch: "[concept] governing principles design criteria" (what laws/principles govern this concept?)

IMPORTANT: Always research:

• When the concept was first introduced (year, original paper/author)
• Sophistication level (basic building block vs cutting-edge)
• Governing principles or laws that constrain the concept (e.g., Goodhart's Law for benchmarks)

3. Extract Semantic Relationships

From research, identify:

• Synonyms: Alternative names (e.g., "cost function" = "loss function")
• Antonyms: True opposites (e.g., "overfitting" ↔ "underfitting")
• Hypernyms: Broader categories (IS-A relationship, e.g., "cross-entropy" IS-A "loss function")
• Hyponyms: More specific variants (e.g., "binary cross-entropy" is a type of "cross-entropy")
• Counterparts: Related concepts often compared (e.g., "precision" & "recall")

4. Create the Note

Read the appropriate template from references/ and create the note:

Frontmatter Templates

Concept (Introduction + Deep Dive)

---
fileClass: Concept
note_type: Introduction  # or Deep Dive