Shall we play a game? Agent vs. Agent

Pierre, writing at ICT en Onderwijs BLOG, rebuilt a three-part agent-to-agent Battleship demo — a referee agent plus two player agents — that Better Stack originally ran with GPT-6 Luna against Claude Haiku 5.5, substituting open source models so he would not be locked into proprietary APIs. The rebuild preserves the referee-and-two-players architecture while swapping the closed models for open source alternatives.

Pierre, writing at ICT en Onderwijs BLOG, rebuilt an agent-to-agent demo he saw in a Better Stack video. The original pitted GPT-6 Luna against Claude Haiku 5.5 in a game of Battleship. Pierre wanted the same three-part setup, a referee agent plus two player agents, but running on open source models, so he wouldn’t be …