Legit question, am I missing anything not playing with agents? Like, what's the killer-app feature vs chatgpt prompt?
Its an order of magnitude difference in what it can do, mostly because it can verify its own work. It can also use subagents which helps a lot. Large tasks I usually have done by subs with the main agent directing. This means the task they can take on can be much larger.
I'm in the same boat as waynesonfire and I'm afraid this doesn't answer the question sufficiently. What do you do with an agent? What's a concrete example vs. typing in a chat box.