From what I understood, it's a coding challenge: the models wrote a player for that specific word game. E.g. https://github.com/rayonnant-ai/aicc/blob/main/wordgempuzzle...

Generally speaking, would you take a conclusion based only an event that happened once?