V1 - Corewar style: only one agent per player is submitted at iteration 0, and all iterations are ran in a single go without any other introductions. This should makes for bigger, more complex agents. Has the advantages not to take weeks to play.

V2 - War of fog: same as with the current rules, except for the fact that the agents of others are black boxes. Maybe a new operand could be introduced to allow an agent to peek into a specific cell.