What's new in Claude: September recap

Summarize your own video

Paste a YouTube link. Peruse reads it and writes the summary.

Anthropic’s September Claude Updates: “Say what you want, and Claude works out how to get there”

Core idea

The speaker explains that Anthropic’s biggest September changes are designed to remove setup friction: instead of choosing a mode or tool and re-explaining your project, you can simply describe your goal and Claude determines the workflow.

Unified experience: Chat + Cowork are now “one Claude”

The speaker says Chat and Cowork are merged, so there’s no longer a separate mode to pick.

A small request stays lightweight (a quick question gets a quick answer), but larger work (like producing a report) can happen in the same conversation without switching contexts.

They also highlight that work now runs in the cloud, so Claude can continue working after you close your laptop, and you can follow along from your phone. Existing Cowork projects remain available where you left them.

Create assets inside any conversation: Docs, Slides, and Design

The speaker introduces a second major change: you can generate a doc, deck, or design from any conversation, without selecting a separate tool up front.

Because these assets build from the context already in the chat, the speaker emphasizes there’s no need to re-brief Claude or start a new conversation just to produce deliverables.

They note that the output isn’t just static imagery: you can directly edit content yourself or highlight a specific part and tell Claude what to change, and Claude will revise only that section.

Claude Docs: collaboration with Claude inside shared documents

The speaker adds that Claude Docs enables teams to work in the same document at the same time as Claude. Team members can mention Claude in a comment to request edits or changes.

They state that Docs, Slides, and Design are available on all paid plans.

Model guidance: Opus 5.5 vs Sonnet 5.5

The speaker attributes these workflow improvements to Claude 5.5, described as a family of models capable of understanding how you want to work.

They recommend:

Starting with Opus 5.5 while you’re still figuring out what you want (exploration and shaping the work)
Switching to Sonnet 5.5 once the direction is clear, because it’s faster and well-suited for rapid iteration (tightening slides, swapping charts, trying new color palettes)

Practical tip: let Claude choose, or use slash commands

The speaker shares a “pro tip”: you can either describe what you want and let Claude choose the best format, or you can jump directly to an output type using commands like /slides (and similarly for docs and designs).

Call to action

The speaker invites viewers to comment on what they’ve created with the new features and which release they’re enjoying most.

Transcript

It's been a busy month here at Anthropic, and September's biggest updates share one common idea: you say what you want, and Claude works out how to get there. Until recently, you had to pick a mode, pick a tool, and then re-explain the project each time you switched. We're going to walk through the two changes that cut that setup, and the new models underneath that made it possible. First, chat and Cowork are now one Claude, so there's no more mode to pick. You ask, and Claude works out the rest. A quick question gets a quick answer, and when you hand it something a little bigger, like, say, putting together a report, it goes off to work in that same conversation. It also keeps going even after you close your laptop, because the work now runs in the cloud, which means you can start off on your laptop and then follow along on the go from your phone. And if you've already been using Cowork, everything you've worked on is right where you left it. The second change: you can make a doc, a deck, or a design in any conversation. So there's no tool to pick either. You do the thinking once, and then ask for one of those design assets, and it builds from what's already in the conversation. There's no more need to re-brief and get a new conversation up to speed. So once you've worked through that report with Opus 5.5 and it's time for a deck, try out Sonnet 5.5. It's faster, which makes it well suited for quick rounds of changes, like tightening a slide, swapping a chart, or even trying a totally different color palette. And it's also important to note that these aren't screenshots. You can fix the slide yourself, or highlight a specific paragraph and tell Claude exactly what to change. Then it edits only that specific part. Claude Docs is also new, and it takes that even further. Your team can now work in the same doc at the same time as Claude, and anyone in it can mention Claude in a comment to make a change. Docs, Slides, and Design are available on all paid plans today. Also, I got a pro tip from the team. You can describe what you want and let Claude pick the format. Or, if you already know what you want, say you want a slide deck, you can type /slides and skip straight to it. The same is true for docs and designs. All of this is possible thanks to the models underneath. Claude 5.5 is a family of models capable enough to understand how you want to work. Start with Opus 5.5 while you're still working out what you want, and switch to Sonnet 5.5 once you know. So now you're totally caught up. Tell us in the comments what you've made with these. We'd love to see. And I particularly love to hear how the community is using these new tools, so please go ahead and leave a comment about which of these releases you've been most enjoying. I can't wait to read them.