GPT-6's most interesting idea may be the interface, not the model.

OpenAI announced a broader GPT-6 rollout alongside 'Intelligent UI,' a feature that lets ChatGPT compose more than plain text. Depending on the task, replies can include diagrams, controls, forms, charts and small interactive experiences. OpenAI says GPT-6 Sol powers the experience for paid tiers while GPT-6 Luna serves Free and Go users.

The company also describes a system of native interface components that can appear progressively as the model responds. This is a design and product shift, not just a new benchmark entry.

For decades, software asked people to learn a fixed interface. AI-generated interfaces propose the reverse: describe a goal, then get a temporary interface built around it. A simple calculation might become a slider, while a comparison might become a table.

That flexibility creates its own challenges. Users need to know which controls merely explore information and which take real action. Dynamic interfaces also need accessibility, consistency, predictable state and a way to verify the claims displayed inside them.

Cortist sees the interface as the next competitive layer. If every answer can become a miniature application, the differentiator may shift from raw model intelligence to usability, trusted data and the quality of judgment behind each interaction.

The right test is not whether a generated tool looks impressive. It is whether real users finish a task more accurately and quickly than they would with a conventional UI—and whether they understand the limitations.