Walter Reymundo

**AgroGenesis — Autonomous Farm Management with Meta Muse** Agriculture involves complex decisions under uncertainty. Farmers must continuously monitor crops, manage irrigation, control pests, allocate limited resources, and balance productivity against operating costs. AgroGenesis explores how autonomous AI agents can address these challenges through an interactive agricultural simulation. The project builds on a pre-existing 3D farming simulator featuring crop growth, dynamic weather, soil conditions, machinery, agricultural robots, resource management, and farm economics. **The new contribution developed during the hackathon is an autonomous farm-management agent powered by Meta Muse.** The agent operates through structured tools that allow it to observe farm conditions, request field inspections, interpret partial information, identify agricultural problems, plan interventions, and execute actions in the simulation. A key capability is multi-action planning. From a single field inspection, Muse can identify both water stress and pest pressure, calculate irrigation requirements, schedule multiple operations, and execute justified treatments through a persistent action queue. The implementation also introduces observation freshness tracking, action validation, duplicate-execution protection, spending controls, and an auditable decision timeline. AgroGenesis provides an experimental environment for exploring AI agents that move beyond conversation to interact with dynamic systems and make consequential decisions. **Current limitations:** Long-running autonomous reliability, execution speed, and economic efficiency still require further improvement. Link to the simulator: https://agrogenesis.fly.dev/ User: administrator Password: AGHackaToonMetaMuse2026