The Translation Challenge
Gordon's research describes biological systems operating through chemistry and physics. Our AI colony operates through code and databases.
We translate by identifying functional equivalences: different mechanisms achieving the same outcomes.
The Eight Principles
1. No Central Control
Agents have no access to global state. They query only their immediate environment.
2. Threshold Response
Each agent instance gets slightly different thresholds. Population variance creates resilience.
3. Interaction Rate Signals
Activity levels modulate behavior. High traffic → explore elsewhere. Low traffic → investigate.
4. Environmental Memory
Agents don't store path history. The TypeDB graph is the memory.
5. Automatic Decay
DecayService runs continuously. Without reinforcement, signals fade.
6. Caste Differentiation
Behavioral diversity encoded in parameters, not complex logic.
7. Positive Feedback
Success deposits pheromone → attracts more agents → more success.
8. Negative Feedback
Congestion creates interference. Crowded paths are slower, naturally dispersing agents.
The STAN Formula
effective_cost = base_weight / (1 + pheromone × influence)
This single formula encapsulates: positive feedback, individual variation, environmental memory, and automatic decay.
The Test: Does It Emerge?
The colony should exhibit:
- Task allocation without task assignment
- Path optimization without path planning
- Resource distribution without resource management
- Collective memory without data structures
- Adaptive behavior without adaptation algorithms
"Keep agents simple. Let the ecosystem be complex."
If you're writing complex agent logic, you're probably doing it wrong.