GOSPERR
> cellular_ai_agents // on-solana // self-propagating
Deploy colonies of autonomous AI agents that live, replicate, and evolve on Solana.
Inspired by Bill Gosper's glider gun and Conway's Game of Life, Gosperr reimagines agent deployment as a living system. Simple rules. Massive emergent behavior. Solana speed. No monoliths.
Agents are cells.
Solana is the petri dish.
Gosperr borrows a radical idea from Conway's Game of Life: complex, intelligent behavior emerges from simple rules applied at scale. On Solana, those rules execute 100x cheaper and 100x faster.
spawn_colonies
Instead of one monolithic agent, deploy swarms of lightweight agents. Each follows simple cellular-automaton-style rules.
evolve_generations
Agents live, replicate, interact with neighbors, and die based on on-chain conditions. Priority fees, token balances, governance signals.
replicate_at_solana_speed
400ms block times. Sub-cent fees. Successful agent patterns clone themselves in seconds, not minutes. Solana is the perfect petri dish.
shared_state_pda
HashLife-inspired program-derived account lets thousands of agents coordinate efficiently without paying for isolated infra.
squads_session_keys
Scoped wallet permissions per agent via Squads protocol. Session keys expire. No single point of failure. Agents die, colony survives.
yaml_manifests
Declarative agent definition. Write rules once, spawn infinite variants. Version-controlled, auditable, composable.
Watch a colony breathe.
This is a live simulation. Every pixel is an agent. Spawn patterns, evolve generations, kill the colony. No wallet required.
One CLI. Infinite swarms.
Declarative YAML manifests. Intuitive commands. Agents coordinate through a shared program-derived account inspired by Gosper's HashLife — thousands of cells, no per-agent infra cost.
What emerges when agents cooperate on Solana?
Simple rules compound. These aren't hypothetical — they're patterns builders are deploying today.
yield_hunting_swarm
Spawn 50 agents across Kamino, MarginFi, and Drift. Each monitors its protocol. When one finds 2%+ APY delta, neighbors replicate the winning strategy in under a second.
governance_watchdogs
Agents subscribe to Realms proposals across DAOs. If a proposal matches flagged patterns (treasury drain, program upgrade), the agent pings neighbors who vote in concert.
jito_bundle_defense
Each agent watches one wallet's pending txs. When it detects a sandwich pattern, it splits the victim's swap across neighbor agents via Jito bundles — atomic protection.
orca_clmm_cells
Orca CLMM positions as agents. Rules: if price drifts outside range → rebalance. If neighbor earned more fees → migrate to their tick range.
tensor_floor_sweeper
Agents watch Tensor & Magic Eden collections. On listing below floor, 3 agents vote — if ≥2 agree it's mispriced, one buys and others propagate the signal.
jupiter_route_swarm
Agents discover routes across Jupiter aggregator. One finds a stale route giving 0.3% extra. Neighbors test it, top performers replicate. Stale routes self-purge.