A lot of research now happens inside a chat window. You paste in sources, ask questions, follow up, ask more. For a quick lookup this is fine. For work that runs longer than a few exchanges, the format starts costing you.
What the transcript loses
A chat is a straight line of turns. That is the wrong shape for synthesis, which is not linear. When you are tying five sources together into a position of your own, you hold several threads at once, revisit earlier points, and notice that turn three and turn eleven are about the same thing. A transcript cannot show you that. It just scrolls.
So the real structure of what you are building stays in your head, and the chat becomes a log of the path you took, which is not the same thing and not much use to reread.
The copy-paste tax
Because the chat is a poor container, you move the good parts out. A useful answer goes into a doc. A quote goes into your notes. A conclusion gets rewritten somewhere permanent. Each of these is a small manual step, and the link between the answer and the source it came from usually does not survive the move.
By the end you have three places: the chat, the doc, the notes. None of them holds the whole picture.
What the work actually needs
Synthesis needs a surface you can lay things out on: sources visible together, your own notes beside them rather than in another app, the freedom to move things as the structure becomes clear, and an AI that reads the surface so asking it something does not mean rebuilding the context each time.
How Graphia handles it
Graphia replaces the chat log with a canvas. Your sources, your notes, and the AI's answers are all on it, positioned where they make sense, staying put. You ask the AI questions the same way, but the answers land on the surface you are building instead of scrolling away, and each one keeps its source attached.
A conversation is a fine way to ask a question. It is a bad way to assemble something.