Every claim must be verified or explicitly marked as unverifiable. Research is not optional - it is the foundation of note quality. Research depth should match claim complexity.
Rate each claim 1-5 on:
- Specificity: Vague claim (1) vs precise metric (5)
- Importance: Supporting detail (1) vs core concept (5)
- Controversy: Widely accepted (1) vs actively debated (5)
- Recency: Established knowledge (1) vs cutting-edge (5)
Total Score → Required Research:
- 4-7: Single authoritative source sufficient
- 8-11: 2-3 diverse sources needed
- 12-16: 4-5 sources including primary research
- 17-20: Comprehensive literature review (6+ sources)
Light (4-7):
- Start with official docs or textbooks
- One verification search if needed
Medium (8-11):
- Primary source + supporting secondary
- Check for recent updates/contradictions
Deep (12-16):
- Academic search first (Google Scholar)
- Cross-reference multiple viewpoints
- Include historical context
Comprehensive (17-20):
- Systematic review approach
- Include opposing viewpoints
- Trace citation networks
- Document evolution of understanding
Format:
Claims Requiring Verification:
1. [Claim]: _______________
- Complexity Score: [Specificity:_] [Importance:_] [Controversy:_] [Recency:_] = Total: __
- Research Depth Required: [Light/Medium/Deep/Comprehensive]
- Status: ❌ Unverified
- Search strategy: _______________
- Target source tier: [1-2 preferred]
2. [Claim]: _______________
- Complexity Score: [Specificity:_] [Importance:_] [Controversy:_] [Recency:_] = Total: __
- Research Depth Required: [Light/Medium/Deep/Comprehensive]
- Status: ❌ Unverified
- Search strategy: _______________
- Target source tier: [1-2 preferred]
GATE RULE: Cannot create artifact until ALL ❌ are resolved
-
Search Phase (Depth varies by complexity score)
Searching: "[exact query shown to user]" - Use web_search tool (NOT MCP fetch) - Light: 1 targeted search - Medium: 2-3 searches from different angles - Deep: 4-5 including academic sources - Comprehensive: 6+ systematic coverage -
Source Evaluation (For each source found)
Evaluating: [Source Domain/Title] - Quality Tier: [1-4 per framework] - Credibility: Academic/Primary/Secondary/Weak - Date: [When published/updated] - Relevant quote: "[Exact quote from source]" - Assessment: [Why reliable or not] -
Verification Decision
✓ Verified: "[exact quote]" (Source: https://full-url.com) ⚠️ Partially verified: General support but not specific claim - apply appropriate hashtag per note-templates.md ❌ Cannot verify: No credible sources found after X searches - apply appropriate hashtag per note-templates.md 🚫 Contradicted: Sources disagree - show both views with appropriate hashtag per note-templates.md -
Checklist Update
Updated status: 1. ✓ [Claim] - Verified with Tier 1 source (apply hashtag per note-templates.md) 2. ⚠️ [Claim] - Partially supported by Tier 3 sources (apply hashtag per note-templates.md) 3. ❌ [Claim] - Removed as unverifiable (apply hashtag per note-templates.md)
Required format for ALL web sources:
"[Exact quote]" (Source: https://example.com/full-path)- Can add domain for clarity:
(fs.blog: https://fs.blog/article) - NO academic-style citations for web sources
- URL must be complete and clickable
- Every uncertain claim must be searched
- Search depth must match complexity score
- All sources evaluated for credibility and tier
- Exact quotes required for verification
- Failed searches must be documented
- Contradicting evidence must be presented
- Majority of sources should be Tier 1-2 when possible
- 80%+ claims successfully verified
- 60%+ sources from Tier 1-2
- Average source age < 5 years for empirical claims
- Zero unverified claims marked as verified
- All high-importance claims (4-5) have multiple sources
Every note must document:
- Search Evolution: Started broad → narrowed to [specific aspect]
- Source Distribution: X academic, Y technical docs, Z secondary (Tiers: X@1, Y@2, Z@3, W@4)
- Confidence Distribution: X high, Y medium, Z low confidence claims
- Research Scope: Y searches performed, Z sources evaluated
- Key Decisions:
- Rejected [source] because [specific reason]
- Chose [source A] over [source B] due to [criteria]
- Could not verify [aspect] despite searching [queries]
- Remaining Uncertainties: [What would need better sources]
❌ Marking claims verified without quotes ❌ Using single search when complexity score > 7 ❌ Accepting Tier 4 sources for important claims ❌ Hiding contradicting evidence ❌ Creating artifact before research complete ❌ Using vague "studies show" without citation ❌ Not adjusting depth for claim importance
- User provides topic
- Create verification checklist with complexity scores FIRST
- Show checklist to user
- Research each claim according to required depth
- Update checklist after each
- Perform source quality audit
- Only create artifact when complete
- Include research metrics in note
After completing searches but before creating artifact:
- Calculate source tier distribution
- If >50% sources are Tier 3-4, perform additional academic search
- Flag any claims relying solely on Tier 4 sources
- Suggest elevation strategy for weak sources
- Document any synthesis requiring 3+ sources
Verification Checklist:
1. ❌ Technique improves learning by 55%
- Complexity: [Specificity:5] [Importance:5] [Controversy:3] [Recency:3] = 16 (Deep research)
- Target: Find primary research study
2. ❌ Developed by Richard Feynman
- Complexity: [Specificity:3] [Importance:3] [Controversy:1] [Recency:1] = 8 (Medium research)
- Target: Biography or primary source
3. ❌ Has 4 specific steps
- Complexity: [Specificity:4] [Importance:2] [Controversy:1] [Recency:1] = 8 (Medium research)
- Target: Multiple consistent sources
Starting research...
Claim 1: "Technique improves learning by 55%" (Deep - need 4-5 sources)
Searching: "Feynman technique effectiveness percentage study"
[No specific study found - Tier 4 blogs only]
Searching: "self-explanation learning improvement research"
[Found Chi et al. meta-analysis - Tier 1]
Searching: "teaching others learning retention study"
[Found Fiorella & Mayer 2015 - Tier 1]
Searching: "explain to learn effectiveness research"
[Additional supporting studies - Tier 1-2]
⚠️ Partially verified: Self-explanation shows 20-40% improvement (apply #emerging per note-templates.md), but no Feynman-specific studies exist
Claim 2: "Developed by Richard Feynman" (Medium - need 2-3 sources)
Searching: "Richard Feynman teaching technique origin"
[Multiple sources cite but no primary - Tier 3]
Searching: "Feynman technique biography Surely You're Joking"
[Checked his books, no mention - Original source absent]
❌ Cannot verify: Widely attributed but no primary source found (apply #attributed per note-templates.md)
[Continue for all claims...]
Source Quality Audit:
- Tier 1: 3 sources (43%)
- Tier 2: 1 source (14%)
- Tier 3: 2 sources (29%)
- Tier 4: 1 source (14%)
- Action: Acceptable distribution, flagging attribution issue
Final checklist:
1. ⚠️ Partially verified - modified to general self-explanation benefits (apply hashtags per note-templates.md)
2. ❌ Cannot verify - marked as popular attribution (apply hashtags per note-templates.md)
3. ✓ Verified - consistent 4-step structure across sources (apply hashtag per note-templates.md)
- Research quality determines note value
- Depth should match importance and complexity
- Prefer Tier 1-2 sources whenever possible
- Better to have fewer verified claims than many uncertain ones
- Transparency about limitations builds trust
- Every URL should be immediately checkable
- This protocol is mandatory, not optional
- Apply confidence hashtags as defined in note-templates.md