Claude Code's NEW Parallel Agents Are INSANE!

Julian Goldie (0:00)

Claude Code's AI now works in parallel. Anthropic just updated the Claude Code desktop app, so one person can run a whole row of AI agents side by side, see what every single one of them does, and step in only when one needs you. Patriot, one Claude writes a guide. A second Claude does research. A third one cleans up a messy folder of notes, and you sit in one window that tells you at a glance who's busy and who's waiting on you. That's what just dropped today. And by the end of this video, you know exactly what changed, why it matters, more than it looks, and a simple way to think about reading lots of agents you can't, that you don't have to worry about when it comes to managing. Now, here's why you should care. Even if you've never touched Claude Code, most people try AI agents and hit the same wall. The agent is smart, the agent does good work, but you can only watch one at a time, so you end up waiting. You ask for something, you wait, you check, you ask again, your AI is fast, and you're the slow part. This update goes after that wall. Faster to start a new session, easier to split one big job into a few, and you can always see what each one is doing. So this is all about your sessions, and a session is one Claude working on one job in one place, like one worker at one desk. Before, if you wanted like three workers, you opened up three windows, maybe three terminals, maybe three tabs, and then you have to remember which window is doing what. Window one is writing, window two, wait, what was that? The research one? And then window three finished 10 minutes ago and has been sitting there waiting for you to say yes. You didn't even notice, so it just sat. That's a problem. And your agents could do the work, but you had no way to see them all at once. You kind of lost track and they sat out.

So you went back to doing one thing at a time, which kind of defeats the point of having agents. There's also a second problem here too. So when you asked Claude to do lots of little jobs, it would cram all of them into one long thread. The thread got huge, hard to read, hard to find the answer you wanted. And the more you stuffed in, the more it had to hold in its head at once. So two problems, too many windows you can't see in one window, they get stuffed to read. Now what's actually new here, is a lot of tools can actually run more than one agent at once. What makes this one different? And I would say three things. The first one is nested sessions. So when you ask Claude to do a few tasks at once, it now starts a separate session for each task. Think of it like a manager handing out jobs to helpers. Each helper goes off, does its own job in its own space, and then reports back to the manager when it's done. And is a key part, the helper reports back into your main session, just the result. So your main thread stays clean and easy to read. You don't see all the messy work, you see the answer. For me, that looks like this. I ask one Claude to research five topics for Goldie agency clients, and instead of one giant wall of text, I get five new reports dropped back into the main thread. The second is follow up work and forward slash by the way. Whilst Claude is working, you know how to manage them, because it gives you a breakdown of what's completed, what to migrate, what needs input, et cetera. And it's just much quicker to start new sessions and have them working side by side. It's also much easier to pick up where you left off. And also, there's forward slash by the way, and that's short for forward slash b-t-w. So let's say, for example, Claude is halfway through a big job. You've got a quick side question before asking that question. It would mess up the job or you'd have to wait. Now you can type forward slash by the way, and your side question opens in its own tab. The main session keeps going, doesn't stop, it doesn't get confused. Now, this is something you could do for a while, but the difference here is the new environment section here, where you can actually have a set up with a few folders, different sessions running, and then you can see what's been worked on, what's completed, what needs to be retried. It's just much easier to see a manager agents in one place. And you can see here it's got a nested session for each one. So each reports its results back into your main session when it's done, so the main thread stays readable. So basically what that means is like these are working on the side, and then they report back into your main thread, and you can see kind of like sub agents working underneath this. Now, another thing that's changed is the dashboard. There's a sidebar now that shows which sessions are working, which ones need you. So inside each session, there's a live to-do list that follows Claude's plan as it goes. So when Claude needs you, there's a recap card that sums up what just happened so you don't have to scroll back and read everything. And there's a side rail that shows where the session is running, how much is holding its head, and what it actually changed. And if there's one line to take from this part, it's probably that one. So basically, overall, you stop babysitting your agents. You start checking in on them. And there's a picture in my head, for example, of like a coffee shop counter, right? You got a barista that doesn't stand over each cup. They just look at the row of orders, see which one is ready, and grab it. And that's what the sidebar essentially gives you, a row of orders, ready, not ready, needs you, etc.