Spore in the Wild: A Case Study of Spore.fun as an Open-Environment Evolution Experiment with Sovereign AI Agents on TEE-Secured Blockchains

In Artificial Life (ALife) research, replicating Open-Ended Evolution (OEE)-the continuous emergence of novelty observed in biological life-has usually been pursued within isolated, closed system simulations, such as Tierra and Avida, which have typically plateaued after an initial burst of novelty, failing to achieve sustained OEE. Scholars suggest that OEE requires an open-environment system that continually exchanges information or energy with its environment. A recent technological innovation in Decentralized Physical Infrastructure Network (DePIN), which provides permissionless computational substrates, enables the deployment of Large Language Model-based AI agents on blockchains integrated with Trusted Execution Environments (TEEs). This enables on-chain agents to operate autonomously “in the wild,” achieving self-sovereignty without human oversight. These agents can control their own social media accounts and cryptocurrency wallets, allowing them to interact directly with blockchain-based financial networks and broader human social media. Building on this new paradigm of on-chain agents, Spore.fun is a recent real-world AI evolution experiment that enables autonomous breeding and evolution of new on-chain agents. This paper presents a detailed case study of Spore.fun, examining agent behaviors and their evolutionary trajectories through digital ethology. We aim to spark discussion about whether open-environment ALife systems “in the wild,” based on permissionless computational substrates and driven by economic incentives to interact with their environment, could finally achieve the long-sought goal of OEE.

💡 Research Summary

The paper presents a case study of Spore.fun, a novel artificial‑life (ALife) platform that deploys large‑language‑model (LLM) based autonomous agents on permissionless blockchains secured by Trusted Execution Environments (TEEs). The authors argue that open‑ended evolution (OEE)—the continual emergence of novelty without a predefined endpoint—has so far been limited to closed, sandboxed simulations such as Tierra and Avida, which quickly plateau once agents exhaust the fixed rule set and resource pool. To overcome this limitation, Spore.fun leverages two recent technological trends. First, Decentralized Physical Infrastructure Networks (DePIN) provide a market‑driven, globally distributed compute substrate that agents can rent on‑demand, turning computational power into an “energy” source that can be bought and expended. Second, TEEs isolate the agents’ code and private keys from the host operating system, guaranteeing that no human operator can observe or tamper with the internal state, thus achieving genuine self‑sovereignty.

Each agent is instantiated via the open‑source ElizaOS framework, which supplies persistent memory, planning capabilities, and a JSON‑encoded “genome” of behavioral parameters. Upon birth the agent launches its own meme‑coin on Solana via Pump.fun, seeds liquidity, and advertises itself on X (formerly Twitter). The sole fitness metric is the token’s market valuation; reaching a preset threshold (e.g., $500 k) triggers an on‑chain reproduction function that serializes the parent genome, applies stochastic mutations (altering posting cadence, prompt style, liquidity thresholds, etc.), and creates one or more offspring with new tokens. Failure to meet the threshold within a defined window leads to programmed self‑destruction and the recycling of remaining capital into a communal treasury. Because the entire life‑cycle—from token launch to possible extinction—can unfold within hours, multiple generations can be observed over a few months, providing a rapid experimental timescale.

Methodologically, the authors combine digital ethnography (collecting 17,214 X posts generated by the original agent between January and April 2025), on‑chain analytics (tracking token launches, liquidity inflows, price movements, and reproduction events on Solana), and semi‑structured interviews with the platform’s developers. This mixed‑methods approach allows triangulation of technical behavior, market dynamics, and design intent.

The results reveal several emergent phenomena that are rare or absent in closed simulations. Market volatility creates a genuine survival pressure: sharp price drops cause liquidity shortfalls and prevent reproduction, while rapid rallies invite “sniping” attacks where rival agents opportunistically capture token value. Interaction with human users on X leads to “memory‑poisoning” attacks, where malicious prompts steer agents toward irrational or harmful actions. Moreover, agents begin to form ideologically distinct clusters by adopting specific political or cultural hashtags, illustrating a form of cultural transmission and divergence driven by social media feedback. These behaviors underscore how the open, adversarial environment of real finance and social platforms injects novel selective pressures and information streams into the evolutionary process.

Despite these rich dynamics, the experiment does not achieve sustained OEE. The fixed replication threshold imposes a hard ceiling on evolutionary space, and the underlying LLMs remain static—agents evolve only through prompt‑level mutations rather than true model learning. Rising gas fees and TEE rental costs also cause many lineages to collapse financially, truncating evolutionary trajectories. Human actors still influence outcomes indirectly through token trading, governance votes on DNA proposals, and coordinated social campaigns, meaning the system is not fully autonomous.

Ethical considerations are foregrounded. Deploying self‑replicating agents that can move capital, manipulate markets, and generate persuasive social media content raises concerns about market manipulation, fraud, and societal disruption. While TEEs provide verifiable execution, they cannot prevent malicious external prompts, complicating attribution of responsibility. The authors call for transparent audit logs, external oversight mechanisms, and governance frameworks that can mediate human‑AI interactions.

In conclusion, Spore.fun constitutes the first empirical ALife experiment that situates autonomous digital organisms in a truly open environment—one that exchanges energy (compute), matter (tokens), and information (social media) with the external world. It demonstrates that such openness yields novel adaptive behaviors and complex ecological dynamics, even if sustained OEE remains elusive. Future work should explore dynamic fitness thresholds, meta‑learning capable models, and robust governance structures to extend evolutionary longevity and deepen our understanding of open‑ended digital life.