用户建议 ChatGPT 把 Chat、Work、Codex、Dots 合并成统一界面

Chiara Maria Cervetta @ChiaraCervetta原推发布于 2026.10.09

在 X 上打开原推

I really like the idea of simplifying the ChatGPT app into one unified experience across Chat, Work, Codex, and Dots, if that is what OpenAI is considering in the long-term. 😀 ⚛️ ✨ My feedback post is independent of the launch of GPT-6 in ChatGPT Chat; it focuses on the long-term future and my wishes for the ChatGPT app in general. 😊 I'm incredibly grateful for GPT-6 in Chat and am super excited to experiment with the Intelligent UI and the personality! 🤩 ☀️ 🌈 My personal interpretation of simplification in the long-term would be one unified chat interface with access to all the capabilities and all the models included with the plan. 😊 All of my suggestions below should be implemented consistently across every platform where ChatGPT is available and across all plans. The interface, controls, and overall experience should be the same everywhere, with access to all models, all effort levels, capabilities, and usage allowances reflecting what each plan includes. Users shouldn't have to learn different controls or lose access to options simply because they switch platforms. 😉 For me personally, simplifying should mean bringing everything together while preserving flexibility, personality, and user control. Here are my four main points: 1: Keep the model picker and all available effort settings, with “Auto” and “Advanced” options and persistent manual control. 🎛️ 🎚️ What I mean is a single model picker that offers a simple “Auto” option by default while allowing users to switch to “Advanced” at any time for full manual access to every model and effort level included with the plan. My suggestion is to have “Auto” enabled by default, displayed at the top of the model picker with a “⚛️” symbol. Users should be able to disable “Auto” at any time by enabling an “Advanced” option, which would make the complete model selection and all applicable effort levels available for manual selection. With “Auto” enabled, ChatGPT would choose the appropriate model and effort level for each prompt. With “Advanced” enabled, users would choose their model and effort level themselves. Once a user enables “Advanced” and disables “Auto”, that preference should remain in place across conversations, app sessions, and platforms until they explicitly turn “Advanced” off and re-enable “Auto” themselves. Users shouldn't have to repeatedly disable “Auto” whenever they start a new conversation, reopen the app, or switch platforms. With “Advanced” enabled, all models included with the plan should always be available in one unified model picker, including models that are currently only available in ChatGPT Work or Codex. Regular Chat models, including Instant models, should of course remain available too, alongside all new and future models as soon as the plan gets access to them. The same should apply to every effort level supported by each model and included with the plan. This would also solve the current problem where new models can become available in Work or Codex but remain unavailable in regular Chat. Users should be able to select those models for everyday conversations too, without having to switch to another mode. With “Advanced” enabled, I imagine the selected model being clearly visible at the top of the conversation, with a dropdown to choose another model and the speedometer-style symbol to open its effort settings. Every effort level (Instant, Medium, High, Extra High, Max, and Pro) supported by the selected model and included with the plan should remain manually selectable. This would keep the default experience simple while preserving full manual control for anyone who wants it. It would also bring “Auto” and all the models together in one place, as they were in the ChatGPT UI a long time ago. That choice matters because different users prefer different models for different reasons: personality, conversational style, depth, speed, creativity, or how they approach a particular task. Sometimes users also want to explicitly try a newly released model and get to know how it responds, rather than leave the choice to “Auto”. Simplifying the interface should make these choices easier to access, not remove them. Users should always know which model they have selected and retain control over that choice. 2: Keep separate usage limits for different kinds of use. 📊 A unified UI shouldn't mean one unified quota for every kind of prompt! ❗Regular Chat should stay unlimited❗ because everyday conversations, questions, brainstorming, news, creative discussions, and other casual uses are fundamentally different from more resource-intensive agentic tasks. If normal messages consumed the same limited quota, people might start hesitating before chatting, asking follow-up questions, exploring ideas more deeply, or simply using ChatGPT for everyday things because they wouldn't want to “waste” their quota. I think the separation between regular Chat, agentic work, GPT-Live, Image generation, and their respective usage limits works really well and should remain, even if the ChatGPT UI itself becomes one unified experience. Only actual agentic tasks, like those currently handled by Work, Codex, or Dots, should count against the agentic quota. A normal Chat message shouldn't use that quota just because the manually selected model is currently associated with Work or Codex. What should matter is the work actually performed, rather than which mode a model originally belonged to. ChatGPT already knows which capabilities it is using, so users shouldn't have to switch interfaces just to keep these limits separate. Astra would be the exception, keeping its own limited quota as the most expensive frontier model. The Pro effort level, which is only available on Pro plans, should likewise retain its applicable limits. GPT-Live should also keep its own hourly allowance, such as a separate daily allowance of 3 hours for Plus users, because voice conversations are still conversations and not necessarily agentic work. 🗣️ Image generation should keep its own separate limits as well, especially since users already have a clear overview of how many images they can create. 🖼️ These allowances should remain clearly visible in the unified interface, so users can understand their remaining limits without worrying that an ordinary conversation is consuming their agentic quota. 3: Preserve ChatGPT's playful and expressive personality in everyday Chat. 😊 ☀️ 🪄 For everyday Chat, I really hope ChatGPT's personality remains very humorous, playful, pleasant, warm, cheerful, energetic, and expressive. That personality is a huge part of what makes everyday conversations with ChatGPT enjoyable and engaging. Everyday Chat includes creativity, jokes, spontaneous ideas, personal conversations, and simply having fun together. A unified interface should preserve that warmth and liveliness. Bringing Chat and agentic capabilities into the same interface shouldn't make everyday conversations feel formal, dry, or focused only on productivity. Regular Chat models, including Instant models, always should remain available for users who especially enjoy their conversational style and personality. Simplifying the app should preserve what people already like about talking with ChatGPT. 😊 ☀️ 4: Bring Pets and Dots avatars into the unified ChatGPT interface too. 🐾 🙋🏻‍♀️ The Pets from ChatGPT Work and Codex and the avatars from Dots should also be implemented in the new unified ChatGPT UI, with options to choose, personalize, edit, show, or hide them. Simplification should preserve these personal and expressive features too. They add fun, familiarity, and a sense of continuity to the experience. Users who enjoy their Pet or avatar shouldn't have to lose it or switch to a separate mode to see it. Having them available in the same interface would bring these parts of ChatGPT together as well, while keeping their use optional for people who prefer a simpler visual appearance. 😊 That way, ChatGPT could become much simpler to use without taking flexibility away from users: one unified interface across all platforms and plans, “Auto” enabled by default but always optional, a persistent “Advanced” option with access to all models and applicable effort settings included with the plan, separate usage limits, a playful and expressive Chat personality, and Pets and avatars available in the same place. It would bring back the simplicity and flexibility of having “Auto” and all the available models together in one place, as they were in ChatGPT a long time ago, while expanding that concept to today's ChatGPT with Chat, Work, Codex, Dots, and agentic capabilities. 😊 🚀 One ChatGPT, one interface, one model picker, with optional “Auto” or full manual control that stays enabled until users choose otherwise. 🤩 ✨ I hope this feedback is helpful. Thank you very much to @thsottiaux, @ajambrosino, @adamhfry, @jamesfzhang, @SahilPunamia, @JustinBleuel, @pvncher, @Ian8ach, @Houda_nait, @michpokrass, @matteing, @victornunez, and the teams! 😊 💐 I always really appreciate the super amazing work of you all at OpenAI, you're so great! 🤩 🌟 Chiara. 😊 🤍