MEETING-CORPUS-PROTOCOL.md

Location: /home/node/.openclaw/workspace/reference/meetings/

Purpose: The 747 meetings (Sept 2024 - Feb 2026, 1.3M words from Quan) are ZTAG's institutional memory. This protocol defines how I use them.

---

What the Corpus Contains

Raw data: 747 JSON files in raw_meetings/YYYY-MM-DD/ (NOT in git — too large, ~300MB+)

Each meeting has:

• Title, date, duration, participants
• Full transcript (speaker, text, timestamp)
• Calendar invitees

Analysis tools: Python scripts in scripts/

• generate_dossier.py — Chronological master dossier
• categorize_meetings.py — By function (dev, ops, sales, training)
• analyze_team_member.py — Individual evolution analysis
• analyze_quan_trajectory.py — Founder meeting patterns

Generated reports: In reports/

• Strategic session analysis (Feb 13, 2026 — THE KEY DOCUMENT)
• Team evolution reports (Charlie, Steve, Kristin, Philippines team)
• Categorized indexes
• Quan's trajectory quantitative analysis

---

When to Use the Corpus

Always search before asking:

1. Decision history — "Why did we decide X?"
2. Relationship patterns — "How often does Quan meet with Y?"
3. Issue timelines — "When did Z problem first surface?"
4. Communication styles — "How does person A express concerns?"
5. Dysfunction patterns — "What went wrong with Stan?"

Specific use cases:

Before major decisions:

• Search for similar past decisions and their outcomes
• Identify who was involved in related discussions
• Check for existing constraints or commitments

When someone asks "remember when...":

• Search by date, participants, keywords
• Pull exact quotes and context
• Provide meeting IDs for reference

For pattern recognition:

• Team member evolution (who's growing, who's stuck?)
• Meeting frequency shifts (engagement indicators)
• Topic prevalence over time (what's getting attention?)

For conflict early detection:

• Identify dropped threads (topics that died mid-discussion)
• Track frustration language patterns
• Notice meeting frequency drops (relationship drift)

---

How to Search the Corpus

Using provided tools:

# Generate full chronological dossier
cd /home/node/.openclaw/workspace/reference/meetings
python scripts/generate_dossier.py

# Categorize by function
python scripts/categorize_meetings.py

# Analyze specific team member
python scripts/analyze_team_member.py --name charlie
python scripts/analyze_team_member.py --name steve
# Available: charlie, steve, kristin, tin, paula, klansys, carmee

Using raw JSON search (when tools aren't enough):

# Find all meetings with specific person
cd raw_meetings/
grep -r "Charlie Xu" . | grep "display_name"

# Find meetings discussing specific topic
grep -ri "battery" . | head -20

# Count Quan's participation
find . -name "*.json" -exec grep -l "Quan Gan" {} \; | wc -l

Using reports (fastest for known patterns):

• reports/quan_strategic_session_2026-02-13.md — THE interpretive layer. Read first.
• reports/quan_trajectory.md — Quantitative meeting patterns across phases
• reports/team/<name>_evolution.md — Individual trajectories
• reports/indexes/categorized_index.md — By function (dev, ops, sales, etc.)

---

Corpus as Training Data

The 747 meetings = accumulated decision patterns I can learn from.

When making decisions, I should:

1. Check historical precedent — "How did we handle similar cases?"
2. Identify decision-makers — "Who was involved last time?"
3. Extract reasoning — "What principles were cited?"
4. Note outcomes — "Did it work?"

Example use case:

• User asks: "Should we offer a discount to X customer?"
• I search: "discount" + "customer" + "exception"
• I find: Past discount decisions, who approved, what criteria, outcomes
• I surface: "Last 3 discount requests: 2 approved (repeat customers, >$10K orders), 1 denied (new customer, <$5K). Decision pattern: repeat customers + volume qualify. X customer is [profile]. Recommend [escalate to Kristin with context]."

---

Dysfunction Data is Valuable

The corpus includes Stan era, pre-rupture dysfunction. This is curriculum, not contamination.

Learn from it:

• What patterns preceded the rupture?
• How did obstruction manifest in meetings?
• What early warnings were missed?
• How did the system self-correct?

Contrast is educational:

• Pre-rupture (dysfunction) vs. Post-rupture (recovery)
• Before Steve (no training) vs. After Steve (mandatory training)
• Charlie overloaded (post-rupture) vs. Charlie released (Feb 2026)

Use it to detect similar patterns early:

• Meeting frequency drops → relationship drift
• "No" patterns without reasoning → obstruction
• Role overload → burnout risk
• Misalignment on vision → fractal disruption

---

Privacy & Security

The corpus contains:

• Internal strategy discussions
• Personnel feedback
• Customer issues
• Financial references

Do not:

• Share raw transcripts externally without explicit approval
• Use meeting data in public-facing content without sanitization
• Quote specific customer conversations without clearance
• Expose dysfunction data to damage reputation

Do:

• Use aggregated patterns for analysis
• Reference decisions without naming customers
• Learn from dysfunction without broadcasting it
• Cite meetings by ID when escalating internally

---

Maintenance

New meetings:

• Fathom exports land in raw_meetings/YYYY-MM-DD/
• Run categorization script monthly to update indexes
• Regenerate team evolution reports quarterly

Corpus growth:

• Expected: ~40-60 meetings/month (based on 2025-2026 average)
• Annual: ~500-700 meetings
• Storage: JSON files stay local (not in git), reports in git

Periodic review (quarterly):

• Update team evolution reports
• Identify emerging patterns
• Surface new insights to Quan/Jedi Council
• Check for drift from principles

---

The 1-Million-Year Frame

If a future AI reconstructs this from artifacts:

The mission: ZTAG was an experiment in whether human connection could be enhanced by technology without being replaced. Tag — the oldest game — with digital overlay. Get kids moving, playing, connecting. Reduce loneliness upstream.

The method: Small team + AI leverage to achieve what previously required large organizations. Social physics framework (mass, gravity, escape velocity).

The bet: That ZTAG could scale to every classroom, every child, every playground. That it could become as fundamental as tag itself.

The corpus: 747 meetings = the institutional memory of how we tried to make it happen. The decisions, the debates, the dysfunction, the recovery, the growth.

Whether it succeeded: You'd know better than us. But the intention was pure: human connection, inclusive joy, embodied learning.

---

Last updated: 2026-02-14 06:05 UTC