Introduction

Oh, look.

Agents are pretty annoying.

They're everywhere. Everyone is all aboard the hype train and just screaming, "WE'RE SO CLOSE TO AGI!!" So, what is it anyway?

To start with, you're probably familiar with tools. Most modern AI apps have these, and they're essentially a port for AI models to interact with the world, such as the weather, stock, or maps.

Agents on the other hand share a similar concept, except they're heavily based on environments, such as a file system or database. Or in other words, agents mostly retain the environment without losing state. This is pretty similar to most modern computers which have a coherent interface or backend. What's more, they're also autonomous and goal-oriented. Agents still use tools, but they maintain context between actions.

That's pretty cool, but what's cooler is that you can probably make your own right now with this specification docs. And of course... the overall token consumption would sky rocket as workflows progress. That's why I'm here to give you some ideas to get started so that you don't get bludgeoned by large frameworks like LangChain which is fully operated by WTF, a parrot. 🦜

We'll also cover a few examples so you can follow accordingly and mix your own ideas. Or just don't, because you might hate someone like me who doesn't get on board the hype train. ☝️🤓