OpenAI is blending GPT-6 and intelligent UI into ChatGPT's everyday chat. When users ask a question, what they get is no longer just a block of text—they can directly receive clickable charts, forms, interactive widgets, and similar content, all without leaving the chat window to open other software. For the first time, the model has grown hands and feet, turning answers directly into interfaces that can be tapped and used.
This move shakes up a product logic that has been in place for years. In the past, mobile internet products organized their functions around fixed app interfaces—if you wanted to track expenses, you went to a budgeting app; if you wanted to compare prices, you opened a shopping tool—whereas intelligent UI breaks this chain: functions are no longer hidden inside each company's standalone apps, but grow on the spot out of conversations as needed. Large models thereby extend from being mere information providers to becoming entry points for task interaction, directly impacting standalone tool and lifestyle apps with single functions. The entry point through which users acquire capabilities has been rewritten, and competition among apps has accordingly spread from a battle over interfaces to deeper levels of data, content, and services.
Following this trend, the hand-built AI app track that previously had a moment of popularity in China is likely to be directly leapfrogged. A batch of platforms that got started by wrapping large models in a shell and building small tools are being forced to transform; the content-plus-tools model of content platforms will also face pressure as traffic gets diverted. Super apps like WeChat and Douyin, which hold mature ecosystems and distribution channels, are instead able to grow differentiated competitive advantages from their own moats—others have to build entry points from scratch, while these apps long ago welded the entry points and scenarios together.
Worth noting is that domestic large models are also exploring another path: connecting the conversation entry point directly to existing services. This happens to complement the intelligent UI's ability to generate interfaces on demand—the former handles complex tasks involving real-time data and transactions, while the latter covers lightweight interaction needs. As the two paths proceed in parallel, the focus of competition among large model vendors will no longer be whose model has more parameters, but who can build the overall capability to organize interactions around user needs and smoothly connect generated interfaces with real-world services.