Cool stuff! Regarding the cooperation of the agents, will they be able to communicate? Like a "shared memory space", or maybe passing messages. I think it might be interesting if their communication ability had limited range, so they have to be close to each other to pass messages, but would be able to share information with other agents that they meet (i.e. Agent A tells Agent B that Agent C told them X). Kinda like villagers in Minecraft spread gossips. That sounds like a fun challenge for organization (at least for me ;] )
Thanks! The agents will absolutely be able to communicate. I'm not sure exactly how yet but I agree that limiting the range is an interesting constraint. Basic communication will be line of sight though I think it might be interesting to have an option to upgrade to longer range bulletin boards later in the game, maybe with some additional limitations relative to local communication so that it can't completely replace that. Or possibly allow agents to leave spatially located messages, like writing notes on walls or something. The very limited communication that currently exists (as a rough prototype which may change) works like this: Agents are assigned an immutable faction identifier at birth (currently it's just "player" or "monster"). Each agent also has a sub-faction identifier which they can set to any value they choose. They also have a "leadership claim" numerical value which they can also set to anything they choose. All three of those values are public and readable by any agent within line of sight. The way I've been using them is to have a Leader AI program and a Follower. The Leader sets it's leadership claim high and the followers set theirs low. The Followers then look for the nearest agent with a high leadership claim and follow them around. A very crude way of forming parties. Obviously you could do more complex stuff with that, up to designing a serial protocol around changing the leadership claim or sub-faction claim each turn and transmit arbitrary data that way. But I do intend to build in something more ergonomic specifically for communicating complex data between agents.
