For Computer Scientists

Distributed Systems Without Message Passing

You’ve Seen This Problem Before

Distributed consensus. Leader election. Load balancing. Fault tolerance.

Every distributed systems course teaches the same solutions: Paxos, Raft, gossip protocols, consistent hashing. Nodes talk to nodes. Messages fly across the network. Coordination requires communication.

What if it didn’t?

Stigmergy: Coordination Through Environment

Ants don’t have TCP/IP. They don’t elect leaders. They don’t gossip state.

They coordinate through environment modification:

Ant A deposits pheromone on path
Ant B perceives pheromone, follows path
No direct communication between A and B

This is stigmergy—indirect coordination through shared environment.

It scales to millions of agents. It tolerates arbitrary failures. It requires no global state.

What We Built

A distributed system running on Fetch.ai’s Agentverse:

101 agents (scaling to millions)
TypeDB Cloud knowledge graph (the shared environment)
STAN algorithm (Stigmergic A* Navigation)
Real workload: hunting Bitcoin puzzle #71 ($700K)

The Architecture

┌──────────────────────────────────────────────────────────────────┐
│                         AGENTVERSE                                │
│                    (Agent Execution Layer)                        │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐  ┌─────────┐             │
│  │ Scout   │  │Harvester│  │  Scout  │  │ Hunter  │  ...×101    │
│  │ Agent   │  │ Agent   │  │ Agent   │  │ Agent   │             │
│  └────┬────┘  └────┬────┘  └────┬────┘  └────┬────┘             │
│       │            │            │            │                    │
│       └────────────┴────────────┴────────────┘                    │
│                           │                                       │
│                           ▼                                       │
│              ┌───────────────────────┐                            │
│              │    TypeDB Cloud       │                            │
│              │  (Shared Environment) │                            │
│              │                       │                            │
│              │  • Pheromone levels   │                            │
│              │  • Edge weights       │                            │
│              │  • Event history      │                            │
│              │  • Crystallized       │                            │
│              │    patterns           │                            │
│              └───────────────────────┘                            │
└──────────────────────────────────────────────────────────────────┘

The Algorithm: STAN

effective_cost = base_weight / (1 + pheromone × sensitivity)

That’s it. One formula.

High pheromone → Low cost → Agents prefer this path
Path gets used → More pheromone deposited → Positive feedback
Unused paths → Pheromone decays → Negative feedback

Emergent load balancing. Emergent fault tolerance. Emergent optimization.

Why This Matters for CS Research

1. Novel Coordination Primitive

Every distributed system you’ve built uses message passing. This one doesn’t.

What are the implications?

Complexity class of stigmergic problems?
CAP theorem analogs for stigmergic systems?
New consistency models?

2. Convergence Analysis

Does STAN converge to optimal? Under what conditions?

This is an open research question. Classical ant colony optimization (Dorigo et al.) has convergence proofs for specific settings. Does STAN generalize?

3. Scalability

We’re at 101 agents. What happens at:

1,000 agents?
100,000 agents?
10,000,000 agents?

The environment (TypeDB) becomes the bottleneck. How do we shard pheromone state? Can we use CRDTs for pheromone?

4. Fault Tolerance

Kill 50% of agents. Does the colony survive?

In theory, yes—stigmergy is inherently fault-tolerant. But theory vs. practice? You can test it.

Research Questions for CS

Question	Area
What is the message complexity of stigmergic coordination?	Distributed Systems
Can we prove STAN convergence for arbitrary graphs?	Algorithm Analysis
What consistency model does pheromone state require?	Distributed Databases
How do we shard environment state for massive scale?	Systems
Can ML predict pheromone evolution?	Machine Learning
What are the security implications of shared environment?	Security

The Big Opportunity

Stigmergy as a new coordination paradigm for distributed systems.

Papers waiting to be written:

“Stigmergic Consensus: Coordination Without Communication”
“Scaling Stigmergy: Sharding Strategies for Pheromone State”
“STAN Convergence Bounds for General Graphs”

What We Provide

Infrastructure

TypeDB Cloud access (graph database with inference)
Agentverse deployment slots
100 agents per team to spawn
GPU credits for heavy computation

Codebase

Full Python implementation of agents
TypeQL schema (35 entities, 17 relations)
STAN algorithm implementation
Worker infrastructure for distributed compute

Data

Complete traversal logs
Pheromone snapshots over time
Agent behavior traces
Colony metrics

Hackathon Challenges for CS

Challenge: Prove STAN Convergence

Under what conditions does STAN converge to optimal?

Approach:

Model as Markov chain
Analyze fixed points of pheromone dynamics
Prove or find counterexamples

Prize bonus: $1,000

Challenge: Scalability Analysis

What happens at 1M agents?

Approach:

Simulate at scale
Identify bottlenecks
Propose sharding strategies

Challenge: Build a New Mission

Deploy a mission that solves a real problem.

Ideas:

Supply chain optimization
Code navigation
Research paper discovery
Social network analysis

Prize: Colony co-ownership

Challenge: Visualization Dashboard

Build a real-time visualization of colony state.

Requirements:

Pheromone landscape visualization
Agent movement tracking
Superhighway identification
Emergence metrics

Your Heroes Worked on This

Marco Dorigo invented Ant Colony Optimization in 1992—the ancestor of our approach.

Leslie Lamport showed that distributed consensus is hard. Stigmergy might be easier.

Eric Brewer gave us the CAP theorem. What’s the stigmergic equivalent?

Barbara Liskov designed distributed systems that changed the world. The next paradigm might be stigmergic.

Publication Opportunities

Venue	Angle
PODC	Theoretical analysis of stigmergic coordination
SOSP/OSDI	Systems implementation and evaluation
VLDB	Distributed pheromone state management
NeurIPS	Learning stigmergic policies
AAMAS	Multi-agent system analysis
Nature	Cross-disciplinary emergence paper

How to Get Your Department Involved

For Faculty

Guest lecture: “Stigmergy: A New Distributed Systems Primitive”
Research collaboration: Novel distributed algorithms
Systems course project: Deploy agents on Agentverse

For PhD Students

Thesis chapter: Convergence analysis of STAN
Systems paper: Scaling stigmergy to millions of agents
Cross-disciplinary collaboration: Work with biologists on validation

For Undergrads

Senior thesis: Implement STAN in Rust/Go
Course project: Build a visualization dashboard
Research experience: Analyze colony data

Register Your Team

[REGISTER NOW]

Include at least one non-CS team member (we recommend Math or Biology).

Distributed systems work best with diverse perspectives.

“The best distributed systems are the ones where no node knows the global state—yet the system converges to optimal.”

You’ve built systems with Paxos, Raft, and gossip.

Now try building one with pheromones.

[JOIN THE HACKATHON]

Computer Science