Not a bad idea - however > Caveman only affects output tokens — thinking/reasoning tokens are untouched. The problem is the thinking. But could help to tune m

Not a bad idea - however > Caveman only affects output tokens — thinking/reasoning tokens are untouched. The problem is the thinking. But could help to tune my system prompt for Kimi.