OpenAI today introduced canvas, a new ChatGPT workspace that opens a separate editing window beside the chat pane. The feature, rolling out in beta to Plus and Teams subscribers, lets users generate writing or code directly in the canvas, then highlight sections to ask the model for targeted edits — adjusting length, tone, or language in prose, or requesting inline code comments, explanations, and bug fixes.

Product manager Daniel Levine demonstrated the tool for TechCrunch, showing how a user could prompt ChatGPT to draft an email, then highlight specific sentences to “make this sound friendlier” or translate the whole thing. For coding, canvas adds a “review code” button that suggests edits the user can approve or decline. OpenAI says canvas will automatically pop open when it detects a longer output or complex task, but users can also trigger it by typing “use canvas.”

The launch puts ChatGPT squarely in competition with Anthropic’s Artifacts, which debuted in June, and with specialist tools like Cursor. As chatbots struggle to complete large projects from a single prompt, editable workspaces let humans fix mistakes without regenerating everything. The question now is whether canvas will keep power users inside ChatGPT — or convince them they needed a dedicated editor all along.