OpenAI is preparing a new step in ChatGPT personalization, according to evidence found in the ChatGPT menu and source code. The update is connected to the code name Sunshine, and it could make ChatGPT better at carrying useful context from one conversation to the next.
The core idea is simple: ChatGPT may become more personal by learning from existing chats. That would shift the product further away from a tool that starts fresh every time and closer to a system that can remember useful preferences, recurring instructions, and details that shape better answers.
What Project Sunshine Appears To Be
The first sign of the update was confirmed in mid-November by X account Jush, who found related options in his or her ChatGPT menu. Web developer Tibor Blaho, who works on the ChatGPT project AIPRM, later showed additional evidence in the source code.
Blaho found functions in the HTML under the code name Sunshine. That is why he called the update Project Sunshine. Based on the source article, the update is not described as a separate app or standalone product. It is presented as a personalization update for ChatGPT.
The reported settings point to a system that can use information across chats. Instead of treating each conversation as isolated, the GPT account could learn over time and use that accumulated context to answer in a way that better fits the user.
The Three Main Memory Features
Three features are described as central to the new personalization settings. Together, they outline how ChatGPT memory could become more useful and more controllable.
- More consistent chats: The GPT account will learn between chats, which OpenAI says should lead to more relevant responses.
- Improves over time: The GPT account should become more helpful as it remembers details and preferences.
- Memories can be customized: Users should be able to shape what the GPT model remembers by sending messages.
The customization part matters because memory can be helpful only if it reflects what the user actually wants the system to retain. OpenAI gives examples of messages that could influence memory, including "Always write in Python" and "Forget everything about my last project".
Those examples show two sides of the same feature. One message tells ChatGPT to keep a standing preference. The other tells it to remove context that should no longer guide future answers. The source also says memory can be disabled in the settings.
That makes Project Sunshine less about adding one fixed behavior and more about adding a memory layer that users can manage. If the feature works as described, ChatGPT could gradually become more aligned with a person’s repeated needs without requiring the same instructions in every new chat.
Why Temporary Chats Matter
Alongside cross-chat memory, OpenAI may introduce temporary chat. The source describes temporary chat as a mode that works like previous chats: information is used only within the chat window.
The temporary chat option is available under the model switcher. Its purpose is to keep the conversation contained. What happens in that chat stays in that chat, rather than becoming part of a broader memory that affects future interactions.
This could be especially useful for teams that share an account but do not always work on the same topics. In that setting, cross-chat memory could create confusion if one person’s work context influences another person’s conversation.
The source also says temporary chats do not appear in the chat history, so the data is not used for model training. That makes temporary chat an important counterpart to personalization: one mode helps ChatGPT carry context forward, while the other keeps a conversation separate.
How This Fits OpenAI’s ChatGPT Strategy
For OpenAI, deeper customization is also an opportunity to sell more accounts. If shared accounts become harder to use because memory crosses between unrelated workstreams, more users or teams may decide that separate accounts make the experience cleaner.
The timing remains unclear. The source says it is not yet known when the new feature will be introduced. That means Project Sunshine should be understood as a reported upcoming update, not as a feature with a confirmed launch date in the source article.
The memory work also fits a broader direction for ChatGPT. The idea of long-term storage for ChatGPT is not new. Back in spring 2023, a developer launched MemoryGPT, described as a cross-chat memory built on top of the API.
The source frames Project Sunshine as part of OpenAI’s larger push to make ChatGPT more individualized and customizable. The logical endpoint of that direction is a more capable personal assistant: a system that can remember what matters, adapt to preferences, and reduce repeated setup across conversations.
The GPT Store Connection
Blaho also says OpenAI is working on an update for the GPT store. That update includes a search function and a stage for the best custom GPTs.
OpenAI is known to be planning a marketplace for GPTs. The launch was planned for November, but it was postponed to 2024 because of the turmoil surrounding the dismissal and reinstatement of OpenAI CEO Sam Altman.
The GPT store and Project Sunshine point in the same general direction: making ChatGPT more customizable. Custom GPTs let users shape specialized assistants, while memory could make those assistants and the main GPT account more responsive to ongoing preferences and work patterns.
For users, the practical question will be control. A more personal ChatGPT can be more useful, but only if it is clear what it remembers, when it remembers, and how to stop memory from carrying across conversations where it does not belong. Based on the reported features, OpenAI appears to be building both the memory layer and the controls around it.