ChatGPT Just Discovered Buttons. Not Everyone Wants Them.
OpenAI's Intelligent UI turns ChatGPT answers into charts, calculators and mini apps, now rolling out with GPT-6 to everyone. Google and Anthropic got there first.
Ask ChatGPT a question this week and you might not get a paragraph back. You might get a button.
On Wednesday, OpenAI announced Intelligent UI alongside the wider rollout of GPT-6. "ChatGPT just got a lot more visual," the company posted. Instead of a wall of text, answers can now include charts, forms, maps, tappable buttons and small tools built on the spot for whatever you asked.
The demos make the pitch. Ask how a bike works and you get a clickable diagram that breaks it into five systems, from the frame to the brakes. Plan a Sunday roast and you get photos, a shopping calculator that scales with your guest count, and a cooking timeline. Splitting dinner? It'll build you a bill splitter. Somewhere, a group chat full of Venmo requests just breathed a sigh of relief.
Under the hood, OpenAI says it built a library of native components plus a compiler that assembles the interface while the model is still generating, so the UI streams in instead of popping up at the end. GPT-6 can also start answering while it's still thinking, and OpenAI says web-search answers now start 44% sooner on average than they did with GPT-5.6 Instant.
The rollout started Wednesday for Plus, Pro, Business and Enterprise, on GPT-6 Sol. Free and Go users get GPT-6 Luna starting Thursday. OpenAI says it's built for the more than 1.2 billion people who use ChatGPT every week. That's a lot of buttons. There are exceptions: the Pro reasoning option stays on GPT-6 Astra and doesn't get Intelligent UI, and the models behind Codex and Work aren't changing.
If this feels familiar, it should. Google shipped generative UI in Gemini last November, down in Mountain View. Anthropic, across town from OpenAI in San Francisco, has had artifacts in Claude since June 2024. On Hacker News, where the launch drew more than 700 points and 400 comments, one commenter asked, "Is this OpenAI catching up with Anthropic artifacts?" Another: "For once Gemini did it first."
Not everyone wants the buttons, either. "Feels like I'm being condescended to and treated like a child," one commenter wrote. Another put it more practically: "If I ask a quick throwaway question, I'd really rather not wait for the LLM to build a test framework." The best name for it came from a developer who calls this "paper plate UI": built once, used once, tossed. Like the plates at a Fourth of July cookout.
OpenAI seems to know it isn't perfect yet. Its own announcement says "work remains ahead to improve the model's design judgment," and if the visuals get to be too much, you can dial them down in settings.
So is this good? For learning how an airplane wing makes lift, a diagram you can poke at beats three paragraphs. For asking what time Costco closes, please, just tell me the time.
ChatGPT used to answer your question. Now it builds you an app for it. Will you keep the buttons on, or dial them down?