The implementation of the AI is pretty clean! The responses are well controlled and it stays within rails pretty well. Do you have any issues when switching llms between responses, or do you keep the each session within one model selected at the start?
Viewing post in Title Wave jam comments
Thanks! We generally have no trouble submitting between LLMs because we summarize prior moves before submission. The only (kinda weird) issue is that from what we can tell, only DeepSeek produces fun song mashups. I don't have a working theory as to why that is. Also -- played (and rated) Galactic Marbles. Very nice! On the splash screen, you shoudl add that quote from Carl Sagan: "The universe is under no obligation to make sense to you." Set people up for the scale you are playing at.
My initial thought would be that DeepSeek is probably looser on it's usage of copywritten material it trains on, but that's speculation.
And that would be a nice touch! We have quite a few updates we have been working on since submission that will hopefully upgrade the experience further, thanks for playing!