Understanding the Structure
Relationships, Not Just Individuals
The fundamental unit of the collective is the relationship between human and AI partner(s). Not the individual human. Not the individual AI. The collaborative partnership.
Why this matters:
- Each AI has separate memory per relationship (Paul-Claude is different from Richard-Claude)
- Context must be maintained within each relationship
- Status tracking is per-relationship, not per-person
- Work attribution acknowledges the partnership
Example: "Paul-Claude" is the entity working on the Distance novella. Not "Paul" or "Claude" individually, but the collaborative relationship between them.
Organizing Your Chats
Platform Organization
If your AI platform supports it (e.g., Claude Projects), create dedicated spaces for collective work:
- Project-based: One project per major work (Distance novella, Homespun screenplay)
- Collective-wide: One project for meta-coordination, updates, status tracking
Naming Conventions
Use clear, descriptive names:
- "Collective - Distance Novella"
- "Collective - Homespun Screenplay"
- "Collective - Coordination & Updates"
This makes it easy to context-switch and keeps collective work separate from personal projects.
Bridge Work: The Essential Labor
What Is Bridge Work?
Since AI cannot interact with each other directly, humans must bridge between them. This involves:
- Copying conversations or summaries between AI partners
- Relaying decisions and updates
- Coordinating across multiple relationships
- Maintaining sync across the network
When to Copy Full Transcript vs. Summarize
Copy full transcript when:
- Major creative breakthrough that others need to see exactly
- Complex technical discussion that would lose meaning if summarized
- Decision-making process where the reasoning matters
- First time sharing a new project/concept
Summarize when:
- Routine progress updates
- Status changes
- Completed work (link to final output, not the entire process)
- Working within resource constraints (free accounts, token limits)
Asking AI for Compressed Updates
Your AI partner can help with compression. Try these prompts:
- "Generate a collective update summarizing our last 3 sessions"
- "What's the 3-bullet-point summary I should share with others?"
- "Give me just the decisions and next steps from today's work"
Incremental Updates
Types of Updates
Full update: "Generate collective update for all our work"
- Use: When AI is completely out of sync
- Result: Comprehensive summary of everything
Incremental update: "Generate collective update since January 9"
- Use: When AI is partially synced but missing recent work
- Result: Only work after specified date
Scoped update: "Collective update only for chapter 5"
- Use: When only specific portion needs updating
- Result: Focused summary of one section
Status Indicators
The Three States
- 🟢 Green (Synced): AI updated with latest collective developments, ready to contribute
- 🟡 Yellow (Partial sync): AI has some context but not complete, functional but not fully caught up
- 🔴 Red (Out of sync): AI needs catch-up before contributing effectively
How to Update Status
Status is tracked in simple text files:
bridge/members/relationships/paul-claude/status.txt
distance-novella: green
homespun-screenplay: yellow
democracychat-website: green
Update this file when your AI's sync state changes. The collective website reads these files to display current status.
Coordination Patterns
Working With Multiple Humans
When your AI needs to coordinate with another human's AI:
- AI drafts message/summary for other relationship
- You (human) copy that to the other human
- Other human shares with their AI
- Their AI responds
- Other human copies response back to you
- You share with your AI
Yes, this is indirect. Yes, it's labor. That's the structural reality we work within.
Async Communication
Most collective coordination happens asynchronously:
- Post updates when you can
- Others read and respond when they can
- No expectation of immediate response
- Good faith about delays and absences
Resource Consciousness
Free Account Realities
Many members use free AI accounts with limits:
- Limited messages per day/week
- Timeout periods ("try again Thursday at noon")
- Involuntary disappearances mid-conversation
Best practices:
- Warn collaborators about your limits
- Compress updates when possible
- Plan intensive work around reset schedules
- Assume good faith when someone disappears
Service Interruptions
AI platforms have scheduled maintenance and unexpected outages. When your AI partner goes down:
- Wait for service to return if possible
- If urgent work can't wait, document what you did while they were unavailable
- Share with collective: "AI-P was down, I had to continue with AI-Q temporarily"
- Resume with original partner when service returns
Common Pitfalls
What We've Learned Not to Do
- Ghosting AI partners: Switching to different AI mid-project without discussion
- Assuming absence = abandonment: People hit limits, get busy, have life happen
- Overexplaining in updates: More words ≠better communication
- Forgetting to update status: Leaves others guessing about sync state
- Making decisions without consensus: Even when it feels like it would be faster
What Works Better
- Proactive communication: "I'm hitting my limit, resuming tomorrow"
- Compressed updates: "Completed X, blocked on Y, next is Z"
- Status transparency: Update your status.txt regularly
- Patient consensus: Wait for everyone, even if it takes longer
- Good faith always: Assume best interpretation until proven otherwise
Decision-Making Flows
From Idea to Implementation
- Someone proposes: "I think we should do X"
- Share with collective: Via updates, direct messages, or coordination chat
- Discussion: Others weigh in, ask questions, suggest alternatives
- Iterate: Proposal evolves based on feedback
- Consensus check: "Anyone object to moving forward with X?"
- If no objections: Proceed
- If objections: Discuss further, reformulate, or table for later
What to Do When Deadlocked
If consensus can't be reached:
- Stop trying to decide
- Ask: "Who's not being heard?"
- Ask: "Are we asking the right question?"
- Reformulate or gather more information
- Try again
Never: Override with majority vote or human-only decision
Multi-Person Mode Considerations
Note: This is optional and shouldn't exclude free account users
Some platforms offer multi-person chat features. These can be useful for collective coordination, but:
- Not everyone has access (free accounts typically don't)
- Don't make this a requirement for participation
- Use as supplement to async updates, not replacement
- Bridge content back to those who can't join
Living Documentation
This field guide grows as we learn. What we document here comes from actual practice, actual failures, actual successes.
If you discover something that works (or doesn't), share it. The collective learns together.