@_jake_ward
i'm trying to figure out the computer
AI
_jake_ward claims the model retains new capabilities without reward hacking when selected parameters are disabled for deployment.