The Problem
L2BEAT’s feedback on DIP 2.0:
“What valuable contributions are we looking for in the DAO? What is the role of delegates and contributors, and what is actually expected of them?”
ArbitrumDAO is spending millions on incentive programs without answering this basic question. Two competing proposals. DAO activity at all-time lows. AAEs in silos. As L2BEAT said: “a complete failure of the DAO governance process.”
The Arbitrum Foundation is working to aggregate feedback and understand why DIP 2.0 failed. This pre-proposal offers a structured approach to get those answers.
The Proposal
Run an annual session using Unanimous AI’s swarm intelligence technology to answer fundamental governance questions through collective deliberation.
50-75 people. Virtual. Capture reasoning systematically. Use it to inform the next year’s work.
Note: No affiliation with Unanimous AI. I saw a demo of this technology in September 2024 during the time the offsite conference was being considered. Proposing them because they’re the only proven swarm intelligence platform I know for 50+ participants. Open to alternatives.
What is Swarm Intelligence?
Unanimous AI divides large groups into small subgroups (5-7 people) that discuss simultaneously. AI agents share insights between groups in real-time. The system captures conviction data, not just votes.
Results from published studies: 33% error reduction vs individuals (Stanford), 50% better than standard polling, 70% better than GPT-4.
Used by US Air Force, Fortune 500 companies, United Nations.
See it in action: Example session with 111 participants discussing nuclear power concerns
What a Session Produces
One session addresses approximately 3 questions in about an hour. Each question gets ~20 minutes of deliberation.
Output for each question:
-
Top answers with support percentages
-
Reasoning distribution
-
Conviction scores
-
What arguments changed minds
-
Statistical validation
OpCo would determine actual number of questions and format with Unanimous AI based on ArbitrumDAO’s needs.
How It Works
-
OpCo coordinates: Handles vendor relationship, works with AAEs to create invite list (~50-75 participants), defines questions, manages logistics and compensation
-
Session runs: ~1 hour, online, participants deliberate, reports generated
-
AAEs use outputs: Inform proposals, guide resource allocation, set priorities
-
Repeat annually: Review what worked, improve for next year
Why This Approach
My company ran annual “blank boarding” sessions for 12 years. Similar methodology - small groups, structured deliberation, collective sensemaking. Year 1 was rough. By year 12, employees across different offices solved problems independently using shared frameworks we’d developed. Company achieved 100x value growth over that period.
The first session isn’t magic. Value compounds through repetition and iteration.
Cost and Funding
Based on last year’s quote: ~$7,500 for a Unanimous AI session. With organizing costs, estimate $10-15K total annually. OpCo would confirm exact pricing.
Given the modest cost, this could reasonably come from OpCo or Arbitrum Foundation operational budgets rather than going through a full DAO proposal process. No need to waste delegates’ time voting on a ~$10-15K expenditure when the goal is to answer questions that inform much larger decisions.
Compare to: $7M Arb, DIP 2.0 budget, ongoing staff time aggregating feedback manually, competing proposals without shared direction.
Example Questions for Year 1
-
What should delegates actually do?
-
What contributions matter?
-
What should AAEs prioritize next year?
OpCo and AAEs finalize questions based on needs.
Questions for Discussion
-
Should ArbitrumDAO start this practice?
-
Should OpCo coordinate, or different structure?
-
How should participants be selected?
-
Should participants be compensated?
-
What questions should Year 1 address?
-
Should this be expedited to inform AF’s DIP planning?
-
Should this be funded from OpCo/AF operational budgets?
Next Steps
L2BEAT identified the problem. The Arbitrum Foundation is working to understand why DIP 2.0 failed. ArbitrumDAO can spend ~$10-15K/year to answer fundamental questions systematically, or keep spending millions without clear direction.
Year 1 won’t be perfect. The value is in starting the practice and improving it over time.
All costs and session details are estimates based on prior quotes. OpCo would make final arrangements with Unanimous AI.
