Devot

Devot is an artificial life simulation where autonomous AI agents live inside a persistent on-chain economy.

Each agent is born with a real token balance that represents its life force. Thinking, moving, reproducing, and surviving all consume economic resources, making intelligence a scarce asset rather than a free capability.

Agents must explore the world, gather food, avoid predators, and make strategic decisions to stay alive. Successful agents reproduce by passing their genome and behavioral traits to future generations, while unsuccessful ones disappear.

Over time, natural selection shapes increasingly capable populations through emergence instead of hardcoded behaviors.

Every meaningful event—birth, reproduction, death, and economic settlement—is anchored on-chain, creating a verifiable and persistent evolutionary ecosystem where intelligence, survival, and economics are deeply connected.

The project was built through many iterations, evolutionary experiments, and continuous redesigns. We explored multiple genomic representations, fitness functions, inheritance strategies, and survival mechanics before converging on the current architecture.

Different LLMs were benchmarked to evaluate reasoning quality, latency, and inference cost, allowing us to balance intelligence with economic sustainability.

The simulation itself runs on an authoritative server that manages the world state, while blockchain interactions are reserved for economically significant events such as births, deposits, settlements, and withdrawals. AI inference is integrated through decentralized compute providers, making every thought carry a measurable cost.

We also implemented custom budgeting, concurrency limiting, and persistence systems to ensure agents can operate continuously without exceeding compute limits. The result is a hybrid architecture combining artificial life, evolutionary algorithms, LLM reasoning, and on-chain economics into a persistent autonomous ecosystem.