Hey guys, Optimizing and testing agents manually is a complete HEADACHE. I still find that manual, gruesome hours sitting there talking to the agent Is the best way to optimize for any obvious edge cases. And for less obvious edge cases. Also, optimizing for stylistic issues like which comma and triple dot should go where for the agent to sound more realistic is just not possible with AI fixing AI prompts. I've watched videos on meta prompting. I've tried building dozens of chat GPT assistants. I tried chat GPT 4.0, I tried chat GPT1o, and none of them are able to optimize the mistakes out of my AI phone caller prompts they just mess it up further it seems like. Glad I have backups. Curious how do you guys optimize your bots to get them from that 80% good to 99% good without hundreds of hours of talking to the damn thing? And even if you do that, you still don't know the perfect way to adjust the prompt for that mistake not to happen + you run the risk of breaking something else in the prompt by adjusting the prompt or adding something.  I want to find a way for AI to fix the prompt but I just haven't been successful at all. They're just too incompetent.