>>105891912
I had to min-max unironically deepseek and switch to v3 because I run around 5 - 10 prompts per message. One being game-master, another analyzing game-master's decision whether its correct, another generating character prompt and another analyzing character prompt whether its aligned to lore. Throw into mix pseudo-thinking out steps and $$$ flows. I'm working on combining local model api with deepseek to cut on costs by running some of prompts on local model instead while using deepseek for ouptut that needs high quality. I can almost feel AGI, AGI loli pussy