This project idea came out of asking the question of what the result would be if one were to implement LLM-based assistant agents that compete over unknown information for their own existence. The key concepts are that there exists a world logic that determines the rules by which the agents define themselves, what information they can communicate and by what constraints, and what are the life / death criteria for each agent.

The agents in the system have an internal state as well, and iterates in time making decisions on what their next actions will be.