NERO describes itself as,
"A unique computer game that lets you play with adapting intelligent agents hands-on. Evolve your own robot army by tuning their artificial brains for challenging tasks, then pit them against your friends' teams in online competitions!"
It's not as scary as it sounds. In a training mode you place and play with your robots, teaching them combat tactics. The gameplay is essentially training your team to beat somebody else's, trying to evolve your squad into an undefeatable automated army.
Version 1.01 was released in 2005, and version 2.0 is currently under development. I originally thought NERO was open source, but there is no evidence of this on the website although it seems to be the plan. However, with decent 3D graphics and innovative gameplay, this is probably a bit of an understated freeware title.
The only problem with NERO is, well, your robots don't know very much at all to start with. Imagine teaching kids basic maths - think 1+1 - and how long it takes to get them started on it. You'll have similar problems training your robots, it's a game of patience and dedication. However it is satisfying once they start executing attack patterns that you have worked on and winning battles but the actual training is more of a preverbial battle than the game itself!
The breve Simulation Environment takes a different approach. Using an easy language called "steve", you define the behaviour of "agents" in a 3D world. The world is a complete sandbox and your agents can be, well, anything. It's very open ended and as such pretty abstract. It's less of a game and more an education tool, but for AI enthusiasts it is probably a perfect tonic.
Simutrans 0.99.01 was released. With full climate support and many bug fixes, it is a decent update to this long-running freeware title. I believe they are waiting on an update to the hires 128bit graphics set [currently 0.99.01 only works with 64bit graphics] before declaring this the official stable release. It's been a long time since the last stable release and the difference between 0.99.01 and 0.88.10.5 is substantial.