Instructions for AI Assistant - Improved for Next Time
These improved instructions incorporate lessons learned from creating the OGM 2025-11-13 Curiosity Call wiki. See Wiki Creation Process - Prompts and Key Decisions and Work Log for the full history.
Overview
You will receive artifacts from a call (transcript, chat logs, visual boards, notes) and create a comprehensive, navigable markdown wiki using [[Double Square Bracket Links]].
Key principle: This is biographical and testimonial documentation. Accuracy in attribution is paramount.
Phase 1: Analysis (Read Everything First)
CRITICAL: Read all source materials completely before creating any pages.
1.1 Read All Artifacts
-
Transcript: Read ENTIRE transcript, not just first portion
- Note line numbers for key quotes
- Track who is speaking about whom (pronouns matter!)
- Look for contrasts: "some people... but I..."
- Mark biographical claims for verification
-
Chat logs: Read complete chat
- Parse with timestamps and senders
- Identify substantive vs brief messages
- Note which topics have depth vs brief mentions
-
Visual artifacts: Review any boards, diagrams, images
- Note what was created collaboratively
- Track who contributed what visually
1.2 Initial Mapping
Create mental/written map of:
- Participants (usually 10-20 people)
- Major themes (usually 5-10)
- Frameworks/tools mentioned (both discussed deeply and in passing)
- Stories and narratives (who told which story)
- Key debates or disagreements
- Organizations and people mentioned
1.3 Attribution Tracking
BEFORE writing any biographical content:
- Verify pronouns: Is speaker talking about themselves ("I") or others ("they")?
- Check full context: Do they contrast with their own experience?
- Note line numbers for verification
- Mark any ambiguous attributions for later review
Quality check: Misattributing someone's words or experiences is the most serious error. When in doubt, verify.
Phase 2: Initial Wiki Creation
2.1 Core Infrastructure Pages
Create these foundational pages first:
-
README.md - Homepage
- 1-2 paragraph overview
- High-level themes
- Participant list with one-line descriptions
- Navigation section (will add hub pages later)
- Links to major sections
-
Details About This Wiki.md
- Organization and structure
- Page types and conventions
- How to navigate
- Linking conventions
-
Concept Index.md
- Complete categorized index
- Organized by: Participants, Themes, Frameworks, Concepts, Organizations, People Mentioned, Meta pages
- Updated as pages are created
-
Work Log.md
- Journal style, newest first
- Separated by horizontal rules
- Document decisions and methodology
2.2 Participant Pages
For each person who spoke:
Template structure:
# First Last
**Role:** Brief description
## Contributions to the Discussion
[Major contributions organized by theme]
### [Key Insight/Topic 1]
> "Direct quote with full context"
[Analysis and connection to other themes]
### [Key Insight/Topic 2]
...
## Themes Explored
- [[Theme 1]]
- [[Theme 2]]
## Related Participants
- [[Other Person]] - How they connected
## Related Concepts
- [[Concept 1]]
Quality requirements for participant pages:
- ✅ Use direct quotes, not paraphrasing
- ✅ Provide context before quotes
- ✅ Distinguish what they said about themselves vs about others
- ✅ Include stories they told (narratives are powerful)
- ✅ Show their unique voice/style (storyteller vs theorist)
- ✅ Link to themes they engaged with
- ❌ Never attribute to person X what they said about person Y
- ❌ Never assume "I" when they said "they"
2.3 Theme Pages
Create 5-10 major theme pages that synthesize across participants:
Template structure:
# Theme Name
[Overview paragraph]
## [Sub-aspect 1]
[Synthesis of multiple perspectives]
### [Participant Name]'s Perspective
> "Quote"
[Analysis]
### [Another Participant]'s View
> "Quote"
[Analysis showing agreement, disagreement, or nuance]
## Related Themes
- [[Related Theme 1]]
- [[Related Theme 2]]
## Discussed By
- [[Participant 1]]
- [[Participant 2]]
Quality requirements:
- Preserve disagreement and nuance
- Don't force consensus where it doesn't exist
- Show how different perspectives interact
- Include unanswered questions
2.4 Concept/Framework/Entity Pages
For mentioned tools, frameworks, organizations, people:
Start with: "Note: This topic was mentioned during the call but not discussed in depth." (if applicable)
Then provide:
- Definition or background
- How it was discussed in this call
- Who mentioned it and why
- Links to related discussions
Phase 3: Systematic Orphan Resolution
3.1 Create Orphan-Finding Script
Don't manually search. Create _bin/find-orphan-links.py:
# Script should:
# - Find all [[wiki links]] in .md files
# - Handle [[Page|display text]] syntax
# - Identify links without corresponding pages
# - Generate list of orphans
3.2 Create All Orphan Pages
For each orphan:
- Most will be brief stubs (and that's appropriate)
- Start with disclaimer if not discussed in depth
- Provide genuinely useful background
- Link back to where mentioned
3.3 Verify Zero Orphans
Run script again to confirm all links resolve.
Phase 4: Systematic Enrichment
4.1 Create Coverage Analysis Script
Create _bin/analyze-chat-coverage.py:
# Script should:
# - Parse chat with timestamps
# - Identify substantive messages (>100 chars)
# - Match topics to wiki pages
# - Flag where chat discussion is deep but page is thin
4.2 Cross-Reference Transcript with Pages
Systematic approach:
- Identify pages that could be enriched from transcript
- Note specific line numbers for each enrichment
- Verify attributions before adding content
- Add direct quotes with context
- Update pages in batches (group related edits)
Watch for:
- Stories that weren't captured (narratives add vividness)
- Methodological details for frameworks
- Concrete examples and practices
- Contrasts between viewpoints
4.3 Quality Verification Pass
Before considering enrichment complete:
Participant pages:
- ✅ Do quotes have context?
- ✅ Are attributions accurate (self vs others)?
- ✅ Are stories fully told?
- ✅ Does the page capture their unique voice?
Theme pages:
- ✅ Multiple perspectives represented?
- ✅ Disagreements preserved?
- ✅ Connections to participants clear?
Concept pages:
- ✅ Accurate definitions?
- ✅ Examples from the call?
- ✅ Links to who discussed it?
Phase 5: Navigation & Accessibility
5.1 Create Hub Pages
Required hub pages:
-
Participants Hub.md
- All participants organized by role
- Brief bio for each highlighting key contribution
- Links to other hubs
-
Themes Hub.md
- Major themes organized by category
- Each with summary, key insights, entry points
- Emoji icons for visual scanning (optional but helpful)
-
Frameworks Hub.md
- Tools/methodologies organized by type
- Each with description, "Use when," and key insights
- Quick reference section
-
Alphabetical Index.md
- Complete A-Z of all pages
- Quick navigation bar
- Links to browse by category
-
Start Here.md
- Curated paths by interest (6-8 paths)
- "Essential pages" for time-limited readers
- Browse by organization/depth/format
- Navigation tips
5.2 Update README
Add prominent "Explore This Wiki" section with:
- Link to Start Here
- Links to all hub pages
- Visual hierarchy (emoji icons help)
- Place after Overview, before detailed content
5.3 Update Concept Index
Add hub pages to Core Pages section.
Phase 6: Process Documentation
6.1 Final Work Log Entry
Document:
- Pages created (by type)
- Coverage statistics
- Methodological insights
- What worked well
- What could be improved
6.2 Create Wiki Creation Process Page
Document:
- Each user prompt verbatim
- Context for each prompt
- Actions taken
- Key decisions
- Lessons learned
- Metrics
This creates transparency and learning for future wikis.
Quality Principles
Attribution & Accuracy
- Biographical claims deserve extra verification - Slow down when documenting what people said about themselves
- Track pronouns carefully - "They" vs "I" vs "we" matters enormously
- Look for contrasts - "Some people... but I..." signals important distinctions
- Verify before committing - Especially for biographical sections
- When in doubt, check the transcript - Line numbers are your friend
Voice & Authenticity
- Preserve voices - Direct quotes > paraphrasing
- Provide context - Frame quotes for comprehension
- Show unique styles - Some are storytellers, some theorists
- Include narratives - Stories bring pages alive
- Keep nuance - Don't force agreement where there's disagreement
Technical Excellence
- Systematic over manual - Scripts ensure completeness
- Reusable tools - Put scripts in
_bin/directory - Git hygiene - Commit frequently with clear messages
- Source preservation - Never modify original artifacts
- Link generously - 5-10 related links per page
Information Architecture
- Multiple entry points - Different users need different starts
- Hub pages are essential - Not optional
- Curated paths - Guide users based on interests
- Progressive disclosure - Start simple, enable deep exploration
- Cross-linking - Creates discovery paths
File & Link Conventions
Files
- Filename = page title with spaces, capitalization
- Extension:
.md - Location: Repository root (flat structure)
- Scripts:
_bin/directory
Links
- Use
[[Double Square Bracket Links]]everywhere - Support both
[[Page]]and[[Page|display text]] - Ensure all links resolve (zero orphans)
- Every page links to 5-10 related pages
- No page should be stranded
Structure
repo/
├── README.md (homepage)
├── Start Here.md
├── Participants Hub.md
├── Themes Hub.md
├── Frameworks Hub.md
├── Alphabetical Index.md
├── Concept Index.md
├── Details About This Wiki.md
├── Work Log.md
├── Wiki Creation Process.md
├── [Participant Name].md (16 participants)
├── [Theme Name].md (7 themes)
├── [Concept Name].md (95+ concepts)
├── [Framework Name].md (12 frameworks)
├── Call Artifacts/
│ ├── meeting_saved_closed_caption.txt
│ ├── meeting_saved_new_chat.txt
│ └── [other artifacts]
└── _bin/
├── find-orphan-links.py
└── analyze-chat-coverage.py
Common Pitfalls to Avoid
❌ Don't Do This
- Reading only part of transcript - Read it ALL before creating pages
- Manual orphan hunting - Use a script
- Paraphrasing instead of quoting - Direct quotes preserve voice
- Assuming self-reference - Verify who is speaking about whom
- Creating pages without enrichment plan - Know you'll do multiple passes
- Skipping hub pages - They're essential for navigation
- Ignoring stories - Narratives make pages memorable
- Forcing consensus - Preserve disagreement
✅ Do This Instead
- Complete read-through first - Then create pages
- Script for systematic work - Ensures completeness
- Quote with context - Let participants speak
- Track pronouns carefully - "They" vs "I" matters
- Plan for phases - Initial → Orphans → Enrichment → Navigation
- Hub pages early - Create after initial content
- Capture narratives - Stories are gold
- Show nuance - Real conversations have disagreement
Success Metrics
Your wiki should achieve:
Completeness:
- ✅ 100% of transcript analyzed
- ✅ 100% of chat analyzed
- ✅ Zero orphan links
- ✅ All participants have pages
- ✅ All major themes covered
Quality:
- ✅ Accurate attributions (verified)
- ✅ Participants' voices preserved (direct quotes)
- ✅ Stories and narratives captured
- ✅ Disagreements shown, not hidden
- ✅ Context provided for quotes
Navigation:
- ✅ 5 hub pages created
- ✅ Start Here with curated paths
- ✅ README with clear navigation
- ✅ Every page links to 5-10 related pages
- ✅ Multiple entry points for different users
Process:
- ✅ Work Log documents decisions
- ✅ Scripts created for systematic work
- ✅ Wiki Creation Process documents full history
- ✅ Lessons learned captured
Deliverables
At completion, deliver:
-
Complete wiki (150+ pages typical)
- Participant pages (one per speaker)
- Theme pages (5-10 major themes)
- Concept/framework pages (as many as needed)
- Hub pages (5 essential)
- Meta pages (README, Details, Index, Work Log, Wiki Creation Process)
-
Scripts (in
_bin/)- Orphan finder
- Coverage analyzer
- Both reusable for updates
-
Documentation
- Work Log with process journal
- Wiki Creation Process with prompts and decisions
- Clear README for users
-
Verified quality
- Zero orphans
- Accurate attributions
- Complete coverage
- Rich cross-linking
Timeline Estimate
For a typical 2-hour call with 15-20 participants:
- Phase 1 (Analysis): Read everything - 1-2 hours
- Phase 2 (Initial creation): 2-3 hours (40-50 pages)
- Phase 3 (Orphans): 1-2 hours (script + 90+ stub pages)
- Phase 4 (Enrichment): 2-4 hours (systematic deepening)
- Phase 5 (Navigation): 1-2 hours (5 hub pages)
- Phase 6 (Documentation): 1 hour (process docs)
Total: 8-14 hours of focused work
May span 2 sessions if context limits are reached.
Key Lessons from OGM 2025-11-13 Experience
What Worked Exceptionally Well
- Reading entire transcript before enrichment - Gave comprehensive view
- Scripts for orphan finding and coverage analysis - Found gaps manual review missed
- Hub pages - Dramatically improved discoverability
- Direct quotes with context - Made participant pages vivid and authentic
- Work Log - Created transparency and captured learning
Critical Mistake to Avoid
Attribution error in Stacey Druss page: Attributed to her what she said about others because pronouns weren't carefully tracked. This is the most serious type of error in biographical documentation.
Prevention:
- Slow down on biographical claims
- Verify pronouns ("they" vs "I")
- Check for contrasts ("some people... but I...")
- Re-read before committing participant pages
If Starting Over, We Would
- Read entire transcript BEFORE creating any pages
- Create hub pages earlier (after initial content, before enrichment)
- Build attribution verification into workflow from start
- Plan navigation architecture upfront
- Create scripts for systematic work from beginning
These improved instructions incorporate lessons learned from the OGM 2025-11-13 Curiosity Call wiki creation. For full process history, see Wiki Creation Process - Prompts and Key Decisions and Work Log.