JarvisTown
Autonomous AI Developer Colony Simulation
Inside the project
01 / 01JarvisTown
The brief
A problem worth
building around.
Creating autonomous AI agents that collaborate meaningfully without scripted workflows is an open research problem. The simulation must give each agent distinct perception, reasoning, and action capabilities while allowing emergent collaboration. Single-binary deployment with no external services requires careful architecture.
Our approach
JarvisTown uses Bevy ECS for entity-component-system architecture where agents are entities with AI "brains". Each agent has perception (what is happening), reasoning (LLM calls with role context), and action capabilities. Agents live in a 2D office environment and collaborate organically when they notice tasks relevant to their role.
The experience
What it lets
people do.
The capabilities that turn the underlying engineering into a usable product.
- 01
Multiple AI agents with distinct specializations
- 02
Role-based personalities and reasoning
- 03
2D office environment simulation
- 04
Task board with emergent workflow
- 05
Artifact generation (code, tests, docs)
- 06
Single binary deployment
- 07
Pathfinding and movement
- 08
Memory of recent events
Project record
What came out of it.
Single Rust binary - no external services
5 distinct AI agents with unique personalities
Emergent collaboration without scripted workflows
Bevy ECS game engine integration
Dioxus desktop UI shell
Leading AI model integration
Under the hood
AI colony simulation with autonomous agents collaborating on software development
AI colony simulation with autonomous agents collaborating on software development
Facing Similar Challenges?
Every business is different, but the problems tend to rhyme. Get in touch and tell us about yours.