amal66 makes chat answers quotable context
A small interaction change could make follow-up questions much more precise without turning chat into a copy-and-paste exercise.
Users can highlight assistant text, add it above the message box as a removable quote, then ask their next question with that passage attached.
- It works in the main assistant, project chats, and tabular-review chat.
- Several excerpts can be collected for one question, then removed individually if they are no longer relevant.
- Long selections are shortened at a sensible boundary and flagged, rather than silently failing.
- Sent quotes remain visibly formatted after reloads and in existing chat history, avoiding a separate data-store change.
The design also limits quoting to assistant prose and keeps an overlong drag within the response where it began, reducing the risk of accidental context.
So what Legal teams reviewing dense answers, tables, or project material should care because it gives the model an unambiguous passage to address.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?