FlyNS

Demo

About

Our project explores what happens when biological intelligence is turned into a network of autonomous digital agents. We use fruit fly connectome data as inspiration for building multiple lightweight agents, each representing an individual digital fruit fly with its own behavior, state, and identity. Instead of treating the swarm as a single AI system, every fly exists as an independent agent that can interact with other flies and with its environment. Each agent is assigned a unique ENS subname, such as `ada.flyns.eth`, which acts as its persistent onchain identity. ENS records can store metadata about the agent, including its wallet address, behavioral state, external endpoints, and other agent-specific information. This makes ENS more than a human-readable wallet name: it becomes a decentralized identity and discovery layer for autonomous agents. The system is designed around the idea of a digital swarm. Individual agents can perceive events, change internal state, communicate, and potentially develop different behaviors over time. By connecting biological network concepts with Ethereum identity infrastructure, we want to explore how thousands of simple autonomous agents could exist, discover each other, and interact through open protocols. The project combines connectome-inspired computation, autonomous agents, Ethereum, and ENS to create an experimental foundation for persistent digital organisms that can live and evolve on the open web.

How it's made

The project is built as a web-based multi-agent simulation that connects connectome-inspired fruit fly agents with Ethereum identity infrastructure. Each fruit fly is represented as an independent agent with its own internal state, behavioral parameters, wallet identity, and ENS subname. We use real fruit fly connectome research as inspiration for the agent architecture, simplifying biological neural relationships into lightweight computational rules that can run interactively in the browser. The frontend is built with TypeScript and Three.js, which renders the flies and their shared environment in real time. Agent state and simulation logic are kept separate from the rendering layer so that each fly can independently process inputs, update its state, and produce actions. For the blockchain layer, we use Ethereum-compatible wallets together with ENS. A parent ENS name acts as the namespace for the swarm, while individual agents receive subnames such as `ada.flyns.eth`. Each subname can point to the agent's wallet and store additional ENS records describing that agent. We also configured resolvers for the subnames so each digital organism can maintain its own independently resolvable identity and metadata. ENS is particularly useful because it gives autonomous agents a human-readable, persistent identifier without requiring us to build a custom identity registry. Instead of referring to agents only by database IDs or wallet addresses, other applications can discover and address them through standard ENS resolution. One notable part of the implementation is that the biological simulation and blockchain identity layers are intentionally decoupled. The flies can continue to run as lightweight local agents, while ENS acts as the persistent external identity layer. This means the architecture could scale from a small visual simulation into a much larger network of independently hosted agents without requiring every behavioral update to be written onchain.