Sometimes during a major conversation flow, I want to start a smaller conversation based a given answer without affecting the main conversation flow, i.e. Digress. Use cases include but are not limited to:
ask to explain a certain term
ask for translation
ask follow-up questions without prolonging conversation
One pain point I discovered using ChatGPT is that conversations tend to become very long and unmanageable. I think by supporting “Digressional Threads”, we can keep the main conversation flow fluent and concise, while giving users the ability to ask elaboration or additional questions.
The end result of this feature should feel like a Slack chat: there are main conversations, and you can have a detailed discussion thread under each conversation.
A key use case for me is programming. When ChatGPT generates a piece of code, I often have follow-up questions about specific parts of it. If I ask these questions in the normal chat flow, the original code moves further up, making it harder to reference and disrupting the readability of the conversation—especially when the follow-up leads to an extended discussion.
A feature like “Digressional Threads” would be very helpful. It would allow follow-up questions to be grouped directly under the relevant response, keeping the main conversation structured and easy to navigate. This would significantly improve usability, particularly for technical discussions or any situation where deeper explanations are needed without cluttering the main chat.
In addition to that, another feature that would greatly enhance usability is the ability to add tags to individual questions or responses. These tags could act as anchors, allowing users to quickly jump to relevant parts of the conversation. A list of these tags could be displayed in a dedicated section (e.g., on the right side of the interface), giving users an easy way to navigate long chats by selecting a tag and instantly being taken to the related part of the conversation.
Both of these features would make working with longer and more complex conversations much more efficient. I hope OpenAI considers implementing them!