Introducing canvas, a new way to write and code with ChatGPT.
OpenAI
OpenAI launched canvas, a new ChatGPT mode for writing and coding side-by-side with the AI. Think Google Docs meets an editor who never sleeps.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI has rolled out canvas, a fresh interface inside ChatGPT built for people who need to actually work on a document or a chunk of code, not just chat about it. Instead of scrolling through a wall of back-and-forth messages, users get a separate window that opens next to the conversation, showing the text or code as a living draft that both you and the model can edit directly.
The pitch here is control. In the old chat-only setup, asking ChatGPT to tweak one paragraph often meant it rewrote the whole thing, forcing you to hunt for what changed. Canvas lets you highlight a specific section and ask for edits just to that piece, or use shortcut commands to adjust reading level, trim length, add final polish, or debug code without restating your entire request. For code specifically, it can review for bugs, add logging, or port a function into a different programming language on demand.
OpenAI built this as a distinct mode you switch into, not a replacement for standard chat. Early access rolled out to ChatGPT Plus and Team users first, with Enterprise, Edu and free-tier users expected to get it as testing continues. The company has framed it as a first version, explicitly asking for feedback, which suggests they know the editing shortcuts and interface choices will get refined based on how writers and developers actually use it day to day.
What's notable is the direction this points to. ChatGPT started as a text box you talk to. Canvas is OpenAI admitting that a lot of real work, drafting essays, fixing scripts, iterating on code, doesn't fit neatly into a chat log. It's a small interface shift, but it signals where they think the product needs to go if it's going to be a genuine tool sitting next to Google Docs and VS Code rather than a novelty search box.
My take — AI-written commentary, not fact-checked reporting
This is OpenAI finally catching up to what tools like Cursor and GitHub Copilot already figured out: chat logs are a terrible medium for actual editing work. I like the direction, but let's be honest, it's a UI patch on a closed model you still can't inspect or self-host, and that tradeoff matters more than a slicker editing pane.
Read more about this at: OpenAI