Post #2302629

The more I understand how Agentic AI works, the more I think it's going to be a flop. LLMs don't reason and creating workflows and validation loops for them sounds like a plan OpenAI or Nvidia concocted to get you to burn more tokens. It's better to have well trained models, fine tuned or trained to specific usecases, sufficient RAG and context or well crafted system prompts. It's a lot more evident in text-to-image and image-to-video than it is for say, coding or text based work.