Cogeneration and Solution Gatling: Accelerating Collective Problem-Solving Through Multi-Agent Collaboration
I’ve been thinking about a fundamental limitation in how we currently approach human-AI collaboration. Most interactions follow a simple pattern: one human operator working with one or several AI agents, picking context and prompts from a single synchronous source. But this creates an unnecessary bottleneck that limits the collective intelligence we could achieve.
What if we could design systems where multiple humans and AI agents contribute context and iterate together on solution development? This led me to conceptualize what I call “Cogeneration” or “Coprompting” - and to develop a practical implementation I’m calling “Solution Gatling.”
The Limitation of Sequential Human-AI Interaction
Current AI collaboration typically works like this:
- Human defines problem and context
- AI generates solutions based on that single perspective
- Human evaluates and refines
- Process repeats with same limited input sources
This approach has several problems:
Single Point of Failure: If the human operator doesn’t capture the problem context well, the entire solution process is compromised.
Sequential Bottlenecks: Each iteration requires the human to process AI output, reformulate, and re-prompt - creating delays and potential misunderstanding accumulation.
Limited Perspective Integration: Multiple stakeholders with different contexts can’t easily contribute to the same solution development process.
Energy Waste: Different team members often run parallel AI sessions on the same problem, leading to duplicated effort and inconsistent solution directions.
Cogeneration: Multi-Agent Collaborative Intelligence
Cogeneration envisions a fundamentally different approach: one interface that accepts multiple context inputs from several humans or agents, all iterating simultaneously with solution-proposing systems.
Core Principles
Parallel Context Capture: Multiple humans can simultaneously contribute their perspective, constraints, and context to the same problem-solving session.
Multi-Agent Orchestration: Different AI agents can specialize in different aspects (creative generation, feasibility analysis, implementation planning) while working on the same problem space.
Asynchronous Contribution: Team members can contribute context and feedback on their own schedules rather than requiring synchronous collaboration.
Interface-Level Optimization: The system design focuses on efficiently capturing and organizing multi-source inputs rather than just processing single-threaded conversations.
Solution Gatling: Practical Implementation
Working with my innovation champion friend Damien Theodorou, I’ve been developing a practical tool to implement these concepts: Solution Gatling.
The Core Insight
The breakthrough insight behind Solution Gatling is this: it’s much simpler for humans to criticize solutions than to describe problems accurately.
Most collective problem-solving fails because getting multiple stakeholders to agree on problem definition is extremely difficult. Everyone has different perspectives, priorities, and ways of articulating challenges. But when you show people potential solutions, they can quickly identify what works and what doesn’t.
How Solution Gatling Works
Rapid Solution Generation: The system generates multiple solution ideas quickly, starting simple (a few words) and becoming progressively richer (descriptions, then mockups, then prototypes).
Multi-User Feedback Collection: Multiple users can quickly rate, criticize, or refine proposed solutions through a simple GUI interface.
Iterative Problem Discovery: Through rapid solution feedback cycles, we indirectly learn what the real problem is - especially valuable for complex collective challenges where the problem itself isn’t well-defined.
Accelerated Iteration: Instead of traditional workshop cycles (propose → wait → design team works → present → feedback → repeat), Solution Gatling enables near-real-time iteration.
Progressive Complexity
Solution Gatling starts with simple solution concepts and gradually increases richness based on collective feedback:
Phase 1: Word-level solution concepts
Phase 2: Sentence-level descriptions
Phase 3: Paragraph-level explanations
Phase 4: Visual mockups or prototypes
Phase 5: Detailed implementation plans
Each phase only advances based on positive collective feedback, ensuring effort is invested in promising directions.
Advantages Over Traditional Approaches
Speed and Scale
Remote and Asynchronous: Team members can contribute feedback whenever convenient, rather than requiring scheduled meetings.
Parallel Processing: Multiple solution paths can be explored simultaneously rather than sequentially.
Reduced Meeting Overhead: Less time spent in workshops debating problem definitions; more time spent evaluating concrete possibilities.
Quality and Clarity
Indirect Problem Definition: Complex problems become clear through what people reject or accept in solutions, rather than requiring upfront articulation.
Collective Requirements Emergence: Instead of trying to extract requirements through metacognition, requirements emerge through iterative feedback on actual proposals.
Reduced Abstraction: Concrete solutions are easier to evaluate than abstract problem statements.
Business Practicality
Less Cloudy Than Traditional Collective Intelligence: Solution Gatling provides concrete, tangible outputs rather than just “increased capabilities.”
Easier to Sell: Stakeholders can see immediate value in rapid solution iteration rather than investing in abstract collaboration improvement.
MVP Acceleration: The test-and-learn cycle for minimum viable products becomes much faster when collective feedback can be gathered continuously.
Technical Implementation Considerations
Interface Design Challenges
Context Aggregation: How do you meaningfully combine multiple different context inputs without losing important nuances?
Feedback Synthesis: How do you aggregate criticism and suggestions from multiple users into actionable solution refinements?
Prompt Optimization: How do you design capture mechanisms that make it easy for people to contribute useful context without requiring AI prompting expertise?
Scalability Questions
Coordination Complexity: As the number of participants increases, how do you prevent chaos while maintaining the benefits of multiple perspectives?
Quality Control: How do you ensure that multiple contributors don’t dilute solution quality through lowest-common-denominator thinking?
Decision Authority: In systems with multiple inputs, how do you handle disagreements and final decision-making?
Connection to Broader Systems
This approach builds on several themes I’ve been exploring:
Collective Intelligence: Cogeneration is essentially collective intelligence applied to human-AI collaboration - leveraging group capabilities rather than individual capabilities.
Compound Thinking: Each iteration builds on previous iterations, with multiple inputs creating exponential rather than additive improvement.
Systems Optimization: Instead of optimizing individual human-AI interactions, we’re optimizing the entire collaboration system.
Current Development Status
Solution Gatling remains conceptual, but Damien and I have been sketching practical implementation approaches. The core technical challenges involve interface design for multi-source input capture and feedback synthesis algorithms that can meaningfully aggregate diverse perspectives.
What excites me most is how this approach could transform not just AI collaboration, but collective decision-making in general. Many organizational challenges stem from the difficulty of getting groups to define problems clearly and agree on solution directions. Solution Gatling sidesteps both issues by focusing on rapid iteration of concrete proposals.
Future Implications
If this approach works as envisioned, it could significantly change how we think about collective problem-solving:
From Problem Definition to Solution Iteration: Instead of spending months defining requirements, teams could spend that time rapidly iterating through solution possibilities.
From Expert-Dependent to Crowd-Capable: Instead of requiring problem definition experts, teams could leverage collective criticism capabilities that most people naturally possess.
From Sequential to Parallel Intelligence: Instead of bottlenecking through individual perspectives, teams could harness multiple viewpoints simultaneously.
The ultimate goal is creating collaboration systems where the collective intelligence genuinely exceeds the sum of individual intelligences - not just through better communication, but through fundamentally different approaches to how problems get solved.