I am looking forward to a feature that allows user to use AI to format the prompt to be more readable. Sometimes students like me will try to query for knowledge when reading materials, and copy-pasting selected text from a external source like PDFs often results in a mess of text, especially when the source contains equations and tables. Such input results in wrong indentation and unreadable texts contained in equations/tables, but can fortunately be recognized by AI (thanks to large amount of PDF sources used for training). I suggest Kagi to add a button next to user's input message, which will try to format the message to be easily readable by human when triggered, using a small language model that is capable to undertaken this action in milliseconds. This will make the whole conversation more worthy to be shared to other users. I strongly suggest Kagi to integrate the Proof&Read feature of Kagi translate into this feature, and allow user to manually choose whether grammar should be checked or not.
For instance, wrong indentation look like this: 
(surprisingly) this whole mess of text can be easily recognizable by LLM. If you ask LLM to reconstruct it in format of Markdown they will give a pretty descent result, even for those smaller-sized model. Currently this can be done with an additional prompt, but still, you know, extra redundant step needed and ugly text is still present in the conversation.