ChatGPT’s code blocks are getting a meaningful UI upgrade: they’re now “more interactive,” with OpenAI highlighting a workflow that combines writing, editing, and previewing code “all in one place.” The announcement, shared via the OpenAI Developers account, frames the change as an effort to make code inside a chat thread behave more like a lightweight workspace—without needing to bounce between chat output and a separate editor just to see what a snippet does.
What changed: code blocks as a mini workbench
OpenAI’s description centers on three capabilities:
- Create and preview diagrams and mini apps in chat
- Review code in split-screen views
- Make changes in full-screen view
That combination aims at a familiar pain point in AI-assisted coding: the “generate → copy/paste → run → iterate” loop. With inline preview and split-screen review, a chat can keep the conversation and the artifact visible at the same time, which is especially useful when iterating on UI fragments, demos, or small utilities.
A few replies picked up on this shift in emphasis. Some developers welcomed the reduced context switching, while others argued that the bigger bottleneck is agents working directly in a real codebase rather than improving a chat-based sandbox.
Diagrams and debugging inside the thread
In follow-up posts, OpenAI Developers noted that interactive code blocks can also be used to preview flowcharts and mermaid diagrams. Another post points to using code blocks “to debug code snippets,” suggesting that the interactive surface isn’t limited to rendering output—it’s meant to support a more iterative debugging loop within the conversation.
A familiar tension: ChatGPT UI vs. dedicated coding tools
The replies also captured a tension that’s been building across AI coding tools: when does the chat UI become “good enough” to keep work in place, and when does it still fall short of what purpose-built tools (and workflows like Codex) handle well? One commenter asked who the feature is for “when Codex is already so good,” while others focused on the practical upside: fast previews, quick split-screen review, and fewer tab hops during prototyping.
The original post: https://x.com/OpenAIDevs/status/2024600394299822096