Have you ever sent a message to an AI agent that spins up multiple sub agents behind the scenes. Each of those sub agents usually go off to do their own work, right. They call tools, they stream tokens, and eventually report back to the main agent to have it composed the final answer.But what does a user see in the meantime. Like just a loading spinner. That's a basic reality of deep agents architecture today.The problem isn't the agent though. It's more the streaming layer. And up until now, our stream hoo ...