I think you’re very interested in why you think the chat interface to LLMs is not great, and why you think your model might be better. But you don’t need people to follow that chain of reasoning for them to see the place you ended up as being interesting.

Show, don’t tell. And start with what the tool does, before you explain why it does it that way.

Might this conversation be an example of the type of thing that you are ultimately trying to address here? Like, could it qualify as a valid use case for using such a tool (if people were serious enough about the topic)?

It seems to me that intuition plays a heavy role here.