Vibeleaderboard
← All Intel
Intel / article

Claude Cowork and chat are now one Claude

Source
simonwillison.net
Date
From the Daily Brief

Anthropic combined Claude's chat interface and its Cowork agentic product into a single assistant. Boris Cherny, an Anthropic engineer, described the change as one experience that carries context across sessions instead of splitting quick questions from long running work. Simon Willison, who tracks Claude releases closely, noted the practical effect: a user can hand off a task and close the laptop, and Claude keeps working, asking for clarification only when it needs input. The merged product also brings Claude's document, slide, and design tools inline, so producing a report or a deck no longer requires switching to a separate surface. The rollout starts with Pro and Max subscribers, with wider availability to follow. For teams building workflows around Claude, the practical change is which surface to design for. A single persistent interface that spans quick lookups and multi hour agentic runs removes the need to route different task types to different products, though it also means one interface now has to handle failure and clarification gracefully across very different task lengths. Anthropic has not said when the merge reaches free tier accounts.

Read the 2026-09-17 Brief →

Context

Anthropic announced on September 16, 2026 that Claude chat and Cowork would share one conversation surface. Users could ask a quick question or delegate longer work without first choosing a separate mode. The company says Claude can use the conversation's existing , skills and connectors, and continue working after the user's laptop closes. Its announcement scheduled the initial rollout for Pro and Max users across web, desktop and mobile.

The change addresses a workflow break: work begun in one surface previously did not carry into the other. A conversation can now progress from discussing a task to having Claude carry it out.

Terms in this piece · Glossary
  • context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
Read the source simonwillison.net
Recommended reads
Comments

Checking sign-in…

Loading comments…