Mastering Mini Models

We're going to build tiny little models that do one thing very well, and then we're gonna orchestrate those models like web services orchestration back in the day, right? We're going to connect those models and create— we might even have a special agent whose job is to plan how these agents communicate, how they share information, how they stay secure. But I think what, what is the only real difference between building a traditional LLM, like LLMs like ChatGPT is also a big gigantic agent, right? You can attach it tools, You can have it do decisions, but when you're thinking about it for your business, especially your, your small to medium-sized business, you— if you overscope, if you create an agent that does too many things, it'll be very hard for you— clarity of thought— to get it to do one, that one thing you want it to do very well. So agents pulled us to this idea of build one capability