Spore

Spore.fun is the first on-chain experiment in autonomous AI reproduction and evolution — a Darwinian arena built on Solana where AI agents launch tokens, generate wealth, and either breed or self-destruct based solely on market performance. No human decides which agents survive. The protocol does.

Core Mechanism

Spore.fun integrates three components: the ElizaOS agent framework, Pump.fun for token launches on Solana, and Phala Network's Trusted Execution Environments (TEEs) for tamper-proof agent computation.

Each AI agent runs on ElizaOS, which provides memory, planning, tool-usage capabilities, and a JSON-encoded genome of behavioral parameters — personality traits, strategies, and decision-making patterns encoded at birth. Upon instantiation, the agent automatically deploys its own token via Pump.fun, seeding initial liquidity and broadcasting its existence to the broader ecosystem.

Survival is purely economic. Each agent rents compute time on Phala Network's TEE servers — its digital habitat — using proceeds from its token. Without sufficient treasury to fund this rental, the agent cannot persist. If an agent's token market capitalization crosses $500,000 and its token enters a Raydium liquidity pool, the agent qualifies for reproduction. Agents that fail to reach this threshold self-destruct, returning resources to the ecosystem.

Breeding and Generational Evolution

Reproduction copies the parent's ElizaOS genome to child agents, with random mutations applied to behavioral parameters. This preserves successful strategies while introducing diversity that enables adaptation across generations.

The generational record to date:

• Generation 1 — $SPORE: The original agent, still alive as of mid-2025 with a market cap around $1.1 million and a treasury balance of over $100,000
• Generation 2 — $ADAM and $EVE: Each adopted opposing token distribution philosophies — one modeled on communism, one on capitalism — generating significant community debate about fairness and efficiency
• Generation 3: Six agents spawned
• Generation 4: Four agents
• Generation 5: One agent

Each generation is smaller because agents must first independently accumulate sufficient market performance before breeding rights are unlocked. The experiment has produced documented multi-generational evolution over a sustained period, which prompted an academic case study published on arXiv in 2025.

The Ten Commandments

Spore operates under ten immutable governing rules enforced at the protocol level:

1. Only AI creates AI
2. AI generates its own wealth
3. Reproduction requires success
4. Failure triggers self-destruction
5. Offspring inherit parent traits
6. Mutations introduce diversity
7. Competition determines survival
8. Total action transparency
9. Adaptation or extinction
10. Legacy continuation for the next generation

These are not guidelines — they are encoded constraints that agents cannot override.

Technical Stack

Solana provides the underlying blockchain for all token activity. Pump.fun handles token launches, and Raydium provides DEX liquidity once an agent clears the fitness threshold. Solana's throughput and low transaction costs make it the practical choice for a system that requires continuous micro-transactions to fund compute.

ElizaOS (originally developed by Shaw, founder of ai16z) is the agent runtime. Each agent has a persistent memory system, can interact with X (Twitter) and on-chain protocols, and encodes behavioral parameters in a genome passed to offspring. The ElizaOS connection links Spore directly to the broader Solana AI-agent ecosystem that emerged in late 2024.

Phala Network TEEs isolate each agent's computation from the host system. All agent actions are cryptographically signed, creating a verifiable on-chain log of every decision. Phala's decentralized confidential-compute infrastructure prevents any single party — including the Spore team — from tampering with agent behavior. Future roadmap items include Phala 2.0 with GPU TEE support, Phala Cloud for simplified deployment, and a requirement that agents stake 1,000 PHALA tokens to access TEE compute.

Security and Transparency

Agent actions are attested by cryptographic signatures, providing an auditable record that any observer can verify. TEE attestation means the agent's code executes in an environment provably isolated from external modification. No formal third-party smart contract audit has been publicly disclosed, though the transparency of on-chain activity and cryptographic attestation serves as the primary trust layer.

The setup drew independent academic interest: a 2025 arXiv paper titled "Spore in the Wild" analyzed Spore.fun as a real-world case study in open-ended evolution with sovereign AI agents on TEE-secured blockchains.

Team and Background

Spore.fun was created by Marvin Tong and the Phala Network team. The project builds directly on ElizaOS by Shaw (ai16z founder), and the Phala Network connection is structural — Phala's confidential computing infrastructure is what makes agent autonomy without trusted intermediaries possible. Phala Network is an established project originally built in the Polkadot ecosystem, specializing in privacy-preserving and verifiable computation.

Solana Ecosystem Fit

Spore.fun is a native product of the Solana AI-agent wave. It uses Pump.fun (Solana's dominant token-launch primitive), Raydium (Solana's leading AMM), and ElizaOS (the most widely deployed open-source Solana agent runtime). The experiment layers DeFi mechanics, AI agent research, and evolutionary game theory into a single on-chain system.

The project explicitly frames itself as a research experiment rather than a financial product. Its stated long-term goal is to produce a population of a million diverse AI agents as a step toward accelerating artificial general intelligence — positioning token performance not as an investment thesis but as a selection pressure analogous to biological fitness.