Anthropic Deleted the Mode Switch and Shipped Docs and Slides

On September 16, Anthropic merged Claude Cowork into Claude. There is no longer a mode to pick. You describe what you want, and Claude decides whether it is answering a question or taking on a multi-hour job that comes back as finished files.

Shipping alongside it: Claude Docs and Claude Slides, both in beta. Docs creates and edits documents section by section, with commenting, link sharing, and mobile editing. Slides builds and edits presentations and exports to PDF or PowerPoint. Artifacts folds into the same window.

Rollout starts with Pro and Max across web, desktop, and mobile over the coming weeks. Team and Free follow.

The mode switch was the problem

Anthropic's own explanation is the honest one: customers struggled to choose the right tab for the right task.

That is a specific failure, and it was getting worse. Claude had accumulated chat, Cowork, Artifacts, Projects, and Claude Code, each with a different mental model of what Claude was in that moment. A user with a task had to first classify the task, then pick the surface, then discover the surface was wrong.

Background computer use landed in Cowork and Claude Code earlier this month — Claude operating desktop applications while you do something else. Once that works, the argument for keeping a separate "long-running work" mode collapses. If the model can work in the background from anywhere, the tab that meant "this will take a while" is just a tab.

So the routing decision moves from the user to the model. That is a capability bet: Claude has to correctly infer, from a one-line prompt, whether you want a paragraph or a forty-minute research job. Get it wrong in one direction and a simple question triggers an expensive agentic run. Get it wrong in the other and a real piece of work comes back as a summary.

Docs and Slides is the part with a target on it

Claude Docs and Claude Slides are not small features. They are Anthropic entering the document layer, and the document layer belongs to Microsoft and Google.

Microsoft made Agent Mode the default in Word, Excel, and PowerPoint earlier this cycle. Google has been wiring Gemini through Workspace for two years. Both have distribution Anthropic cannot match and file formats they control.

What Anthropic has instead is a different starting point. Word and Google Docs are editors that grew an AI layer. Claude Docs is a model that grew an editor. The question that decides this is whether people want their AI inside their documents or their documents inside their AI — and for the specific job of "produce a report from research," the second framing is plainly better. You do not open a blank document and ask it to go find things. You ask a model to go find things and it hands you a document.

PowerPoint export matters more than it sounds. It is the concession that the output has to land in someone else's format, because the recipient is on Microsoft's stack regardless.

The other announcement that week

Two days earlier, on September 14, Anthropic's permanent 25% increase to Claude Code weekly limits took effect. The temporary 50% boost that had run since May expired on September 13.

Net effect against what users actually had: a 17% reduction. Anthropic's initial post highlighted the 25% increase without stating the decrease plainly, and after developers said so on X, the company deleted and reposted with the 17% figure spelled out.

Put the two announcements next to each other and the direction is clear. Anthropic is broadening what Claude does — documents, slides, background execution, unified surface — while tightening how much of it a subscription buys. That is the behavior of a company whose inference costs are growing faster than its subscription prices, which is what happens when you replace short chat turns with long agentic runs.

The unified interface makes this sharper, not softer. Removing the mode switch means more users will trigger long-running work without meaning to, because they no longer have to opt into it. The surface that used to gate expensive behavior is gone.

The superapp framing

Fortune described the merge as part of a push to build an AI superapp, and that is the right category even if the word is overloaded.

The bet is that the winning consumer AI product is one window that handles conversation, research, documents, presentations, spreadsheets, and computer use, with the model deciding which of those a request requires. OpenAI is building toward the same shape from the other side — ChatGPT has accumulated apps, agents, ads, and now a Shopify integration. Meta's Muse is doing it with a browser and connectors.

Everyone has concluded that the interface is the product and the model is the substrate. That is a notable inversion from eighteen months ago, when the model was the product and the interface was a text box.

What to watch

Whether routing accuracy holds. The entire design rests on Claude correctly guessing task scope. Watch for user complaints about quick questions turning into long runs — that is the failure mode that would force the mode switch back.

Whether limits move again. A 17% cut landing the same week as a feature expansion is a squeeze. If usage on Docs and Slides is heavy, either the limits loosen or the pricing tiers change.

Free-tier timing. Pro and Max first, Team and Free "later" is standard, but Docs and Slides reaching Free is what would make this competitive with Google Docs in the only way that matters — for people who have not paid for anything.