Big Idea

The argument in one line.

The rebuilt Claude Code desktop app is no longer just for coders, and its projects, side chats and split view let one person run several AI tasks at once without wrecking context.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You use Claude Cowork every day and haven't tried the Claude Code tab because it looks like a tool for programmers.
You juggle several Claude chats for one larger project, like a YouTube video with separate thumbnail, outline and editing threads.
You have a pile of old skills and CLAUDE.md instructions and suspect they're burning context for no reason.

SKIP IF…

You already live in Claude Code and run nested sessions and slash commands daily.
You want a deep tutorial on one feature. This is a fast tour, and the creator points to separate videos for dashboards, design and motion.

TL;DR

The full version, fast.

Anthropic rebuilt the Claude Code desktop app, and it is now the place new Claude features ship first. The layout adds a live to-do list plus environment, context, connector and output cards. Projects, or nested sessions on team plans, let one coordinator chat spin up threads for separate tasks. Side chats answer questions without touching the main context, split view shows several chats at once, and the docs, slides, design, dashboards and motion tools publish as artifacts. Four slash commands help: output style, doctor, skill doctor and advisor.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 00:58

01 · Why switch from Cowork

Intro and the case that Claude Code is not only for coders and gets new features first.

00:58 – 02:55

02 · New layout and cards

To-do list, environment, remote control, context, connectors and outputs cards, plus chat sections.

02:55 – 06:30

03 · Projects and nested sessions

One coordinator chat spins up threads for separate tasks, with shared memory on projects.

06:30 – 07:47

04 · Side chats

Persistent side chats read the main chat without adding to its context.

07:47 – 08:16

05 · Split view

Show several chats on one screen by dragging from the sidebar.

08:16 – 10:13

06 · Docs, slides, design, dashboards, motion

Anthropic's artifact-based office suite with live dashboards and code-built motion graphics.

10:13 – 12:30

07 · Four slash commands

Output style, doctor, skill doctor and advisor.

12:30 – 13:40

08 · Mods and wrap-up

Build your own mods, like a brain dump cleaner, and grab the PDF guide.

Atomic Insights

Lines worth screenshotting.

Claude Code is the same engine as Cowork with more features, so switching is not a coder-only decision.
New Claude features are launching in the Code tab first, which is a practical reason to stop using Cowork.
A context card shows what is filling your context window and gives you a one-click compact button.
Projects add a coordinator layer above sub-agents, and unlike sub-agents you can talk directly to each thread.
Projects keep a shared memory across threads, while nested sessions only get a brief and prompt from the main chat.
Project threads can run in the cloud with your laptop closed, but nested sessions only run locally.
One chat per task is a good habit, and Claude now offers to start a new thread when you go off-topic.
Side chats can read the whole main chat but never add to its context window.
Side chats are persistent and can send messages back to the main chat, unlike the temporary slash-by-the-way chat.
Dashboards pull live data from your connectors every time you open them.
Claude Motion builds animated video in code without a video model, which makes it far cheaper.
Running doctor on one account found 312 problems across 404 files, mostly instructions newer models don't need.
Skill doctor shows what each installed skill costs in context, and deleting unused ones saved about 20,000 tokens.
Advisor lets a cheaper model ask a stronger one for help only at key moments in a task.

Takeaway

Run big work as threads, not one chat

WHAT TO LEARN

Splitting work into coordinated threads, side chats and split views keeps context clean and lets several tasks move at once.

Claude Code and Cowork share the same engine, so the Code tab is a fuller version of the same tool rather than a coder-only product.
Watch the context card and compact before the window fills, so the model keeps performing at a high level on long chats.
Give each task its own thread and let a coordinator track the whole project instead of cramming everything into a single overloaded chat.
Use side chats for questions and reviews that shouldn't pollute the main context, since they read everything but add nothing.
Audit your setup with doctor and skill doctor, because outdated instructions and unused skills quietly consume tokens every session.
Pair a cheaper model with advisor so the stronger model only weighs in on the moments that matter.

Glossary

Terms worth knowing.

Nested sessions: Chats started from inside a main chat. The main chat can message them, check their work and get reports back.
Compact: A command that summarizes the chat to free up space in the context window so the model stays sharp.
Side chat: A persistent chat that can read the main chat up to that point without adding anything to it.
Artifact: The canvas-style output for docs, slides, designs, dashboards and motion that you can edit with comments and publish to a link.
Output style: A setting that controls how Claude words its answers, such as short and simple instead of highly technical.
Mod: A add-on that changes how the Claude Code desktop app looks and behaves, built by describing it to Claude.

Resources

Things they pointed at.

10:40toolClaude Code /doctor

11:40toolClaude Code /advisor

Quotables

Lines you could clip.

05:20

“Instead of managing five chats yourself you can talk to one coordinator who runs a small team of agents or sessions for you.”

clear one-line pitch for the projects feature→ TikTok hook↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

Entropic just rebuilt the Cloud Code desktop app entirely and launched a bunch of powerful new features like projects, threads, mods, side panels, multitasking, and more, which can all really improve the way you work with AI. So in this video, I'll show you every new feature and exactly how to use them. And I'll explain why you probably want to switch to Cloud Code if you still use Cowork.

Now, if you don't know me yet, I'm Ben. I'm a three -time founder and now run an AI agency, an AI second brain app, and one of the largest AI communities for professionals and business owners, where I teach how to use AI to actually... actually automate work or build an AI business.

So the first thing I want to quickly cover is the new Cloud Code desktop layout, which got a big improvement. Now, until recently, I mostly used Cowork because I just preferred the layout and the UI. But I strongly believe that if you still use Cowork today, you really want to consider switching to the Cloud Code tab because first of all, they've improved the layout a lot.

Second, it also seems to be the place where Entropic is launching their newest features first. And third, because some of these new features are really powerful, it can really improve the way you work with AI. And for anyone who's hesitant, it's really not only for coders.

It's essentially the same tool as CodeWork in the back, just with much more functionalities. For example, Cloud Code can work with any file on your computer. It has access to CLIs, and it has now access to all of these new features that I'll cover in this video.

And I would argue that with this new layout upgrade, it looks a lot less intimidating too. So starting with the layout upgrades. Now, one of the big reasons I used to like working in CodeWork besides its UI was its site panel, because it gives you a clear plan of Cloud's to -do list.

all the connector skills and files it uses in every chat. But now Cloud Code has exactly the same and even more. So every time Cloud makes a plan, you now get a to -do list inside of the chat and you can watch it tick off each tasks as it progresses.

Next to the chat, you'll also get a new set of cards. So the environment card, for example, shows you which folder in the chat runs and whether it's a local or a cloud chat. We also have a remote control switch, which if you turn it on, lets you continue a local chat in a cloud.

So you can keep working on that task from your phone. The new context card shows how full your chat is and what's actually taking up space in your context window with a shortcut button to compact your chat. And that basically means Clod summarizes the chat to free up space every time it reaches a certain amount of tokens inside of the context window.

So you can keep talking to Clod while it's still at high performance. Then there's also a connectors card and an outputs card showing every file the chat made or used. And lastly, you can now sort your chats.

into sections, which basically allows you to group chats together by category. To create a section, all you do is you click on the three dots next to any chat, go to move to new section, name it, and then create the section. Then you can easily move any chat into that section by just dragging and dropping it.

And for me, this has honestly been amazing to organize my chats because their earlier grouping feature was just not good. And it always took me ages to find my old chats, which brings me to the next feature, which is probably my favorite, the new cloud code project. nested sessions depending on your plan now entropic is rolling these new features out slowly so it might look a little bit different for you but projects are now on pro and max and on a team plan you'll see nested sessions instead but the idea is essentially the same it essentially allows you to work on larger projects more efficiently the idea is simple one main chat or project can start other chats inside of it also called nested sessions they can send the messages they can check their work and report back to you so instead of managing five chats yourself you can talk to one coordinator or orchestrator who runs a small team of agents or sessions for you now this is really useful if you work on larger projects in cloud for example when i prepare for youtube videos i usually have one chat to ideate on thumbnails and titles another chat where i go back and forth figuring figuring out the outline and another one for the visuals and editing because of course there are multiple tasks in this larger project of creating a youtube video now the problem with cloud or the old

projects was that these different chats had no context on what happened in the other ones. So even though a change in one of them should affect what happens to the other it just doesn't know. It also means you have to switch sessions constantly and of course if you do all of these tasks in one chat it gets messy quick and the context window gets overloaded fast and you will lose performance.

So with these new nested sessions or projects I can just start a new project and talk to the main coordinator or orchestrator agent which can spin up multiple threads or nested sessions for each of these tasks. So one for thumbnails and titles, one for outlines and one for visuals and editing.

I can then iterate on all of them from the same chat without overloading the coordinators context window. I can even still iterate inside of each of the specific sessions too. Meanwhile, the main coordinator still knows exactly where we are in the process in each of the threads.

So it just makes it a far more efficient way to work on and organize larger tasks you do with AI. Now you might think, didn't we have sub agents for this? but these threads and sessions can still spin off sub -agents on their own.

So essentially with these new projects and nested sessions, we get one extra layer of agents above them, a coordinator or an orchestrator. And of course, you can't iterate directly with sub -agents, which you can do with these threads or nested sessions. So if you have a bigger project, I highly recommend starting to use this because it's extremely useful in my experience.

Now, if you have access to the new projects, you can just start a new project and it will start spinning up these threads automatically. And if you're on a team plan and you don't have access to the new projects yet, you can just open a blank chat and ask it to work on a bigger project. And Cloud will now automatically offer to spin up different threads for you, or you can just tell it to spin up threads.

And whenever you ask for something unrelated in the middle of a chat, it will also offer you to start a new thread, which is a good habit to have anyway, one chat per task. Then all of the threads it spins up show up in a little dropdown underneath your main chat, where you can still go in and iterate with that. Now, there are a couple of differences between projects and nested sessions.

As of today, projects keep a shared memory, so every thread adds to it and draws from it, while nested sessions get a brief and a prompt from the main chat instead. And project threads can run in the cloud, so they can keep going with your laptop closed, while nested sessions only run locally on your computer. But Entropic is slowly making projects available for all plans, and we should get them in a couple of weeks.

Now, very quickly, before I show you the next feature, if you enjoy this type of content, you'll really enjoy my AI accelerator. where you'll get access to my 30 -day cloud course that takes you from beginner to pro in less than one hour a day. You also get access to a full course on growing an AI business based on everything me and my team have learned over the last three years, unlimited one -on -one live AI consulting with AI experts and a big library of skills, templates and resources across all business departments.

So definitely check it out if you wanna dive deeper into AI in the second link in the description below. Now the next new feature is also very useful and it's called side chats. Now, if you've used cloud for long tasks, you probably know - the feeling when cloud is halfway through its to -do list but you have a question and you don't want to interrupt or derail cloud so we now got a fix for this you can now just open up a side chat by clicking on the three dots on the top right and then selecting side chat you'll notice that a side chat pops up with its own chat window and you can open as many of these side chats as you want now each side chat can read the whole main chat up to that point so you can ask what it's doing how long it might take or have it review the work so far without adding adding any to the main chat.

Now this is also a great feature because you can ask the side chat questions that are not entirely related to the main task you're doing and it never ends up in the main context window. Now this is sort of a mature version of the slash by the way if you're familiar with it which would open up a temporary side chat but now these side chats are persistent which means they actually get saved and these side chats can even send messages to your main chat.

I've used this for example when I want the side chat to review the output of the main chat. Now the next one is multitasking or split view with Cloud Code, which fixes the hassle of hopping between different chats all the time.

Now with this new split view, you can have several chats on one screen at the same time. You can simply drag chats from the sidebar into your current chat and you should see an option to add a split view. Or you can click on the three dots next to any chat in the sidebar, select open in and then select split view.

Again, really useful so you can work on many tasks at once and multitask between different chats. Now the next one isn't brand new, but it's so useful. it is still worth covering if you haven't tried it yet, which is Cloud's new docs, slides, design, dashboards, and motion features.

Now, Anthropic is basically building its own Google Docs, Google Slides, a design tool, Looker Studio, plus a video creation tool, all inside of Cloud. Now, every output of all of these new features is in what they call an artifact, which is sort of like a canvas that allows you to really easily edit and iterate together with the comment feature.

You can also easily publish them through artifacts as a live page. on a link so you can share it with your team or with anyone else who has the link so starting with dashboards which launched very recently now these new dashboards can pull data directly from your connectors and now show live data meaning every time you open it it pulls the latest data from your connectors to show you the most up -to -date metrics i'll do a full video soon on this new dashboard feature because it's very powerful then their new slides feature lets you create animated presentations and slideshows straight from cloud and download them in ppt format or present them right from Cloud.

Now, Cloud Docs is essentially Google Docs with tabs where your team and your AI agents can collaborate on documents in real time. I now use this to prep all my YouTube videos. Cloud Design gives you a Figma -like board or canvas, which is now my go -to for any design work I need for social media, websites, et cetera.

And because of the canvas, I can easily change copies, leave comments, and have Cloud fix things. I did a full video on Cloud Design, so I'll make sure to link it in the description below too, if you want to learn more about this feature. And finally, Cloud Motion lets you make animated videos or motion graphics all in code.

It can create really impressive videos without a video model. It's just using code to build these videos, which means it's far cheaper than a video model. And again, because we get the canvas, you can easily iterate and improve on videos fast.

Now, I also did a full video on Cloud Motion and its use cases, which I'll also link in the description below if you want to learn more about this feature. Now, Cloud Code also launched a few new slash commands, and this is another reason you really want to consider switching from co -work. to code if you haven't yet because cloud code has far more slash commands now a slash command is basically a shortcut you type starting with a slash of course and there are four specifically that are really worth knowing about now the first one is slash output style and with it you can basically control how cloud gives you outputs for example sometimes cloud gets really technical so we created our own output style which keeps cloud's answers short and simple now if you're interested in it i'll also put it in the free resources in the first link in description below but if you want to create your own output style you can just have cloud create one and then add it as an output style and invoking it through the slash command now the second one is slash doctor now many people have lots of skills and a cloud .md that are outdated or not optimized for newer models now this command fixes that by automatically reading your cloud .md files which are the first instructions cloud reads in every chat plus all of your skills and all other settings for outdated or conflicting instructions which might may not be relevant for a 5 .5

models anymore. So in my case, it analyzed 404 files. It found 312 problems like instructions that aren't needed anymore for Opus 5 .5 and are consuming unnecessary tokens.

It's just a nice one to run sometimes to clean up your Cloud account. Now, the third one is slash skill doctor. Now, every skill you install takes a bit of Cloud's context.

It basically means it's working memory, even when you never use that skill. So the skill doctor command shows what each skill costs and how often you actually use it. Flags the ones you never really...

use and makes recommendations on which ones you probably want to delete now this one for example saved me about 20 000 tokens which were being consumed by skill files i never actually used and lastly slash advisor which isn't that new but it's one of my favorites especially with the neo 5 .5 models so if you run a cheaper model like sonnet or haiku with slash advisor clot can ask a stronger model like opus for advice at key moments in a task so you get the stronger models judgment where it matters without for it the whole time.

For example, I use this regularly when I use Sonnet to clean my inbox. I use it together with Slash Advisor. And whenever there's something critical, it uses Opus 5 .5 to help draft a response that's important.

And the last one is mods, which is one of my favorite new features. A mod basically changes how Cloud Code looks and behaves. And it essentially allows you to build features inside of the Cloud Code desktop app yourself.

For example, one that I built is a brain dump mod because I do a lot of long voice. brain dumps when prompting cloud, but voice brain dumps often contain errors. They jump around, they're unstructured and sometimes even contradict themselves.

So this mod detects when I've used a long voice brain dump and automatically turns that into a clear structured prompt before cloud even processes it. Now you can build these mods really easily by just telling clouds to build your mod and describing what it needs to do. I just did a full video covering nine cloud mods that I now use on a daily basis, and I'll show you exactly how they work and how to build them yourself in that.

video. So if you want to learn more about this, I'll also make sure to link it in the description below. Now I've put the prompts and a summary of every feature in a PDF guide, which you can download for free in the first link in the description.

And if you want to get access to my 30 day cloud course that takes you from beginner to pro in less than one hour a day, plus unlimited one -on -one live AI consulting and hundreds of free resources across all business departments, you can check out my AI accelerator in the second link in the description below. Thank you so much for watching.

And if you got any value out of it, I highly appreciate a like and a subscribe. It really does help me. Thank you.

The Hook

The bait, then the rug-pull.

The Claude Code desktop app was rebuilt, and Ben says it's now the better home for anyone still on Cowork. He tours every new feature in about 14 minutes.

Storyboard