The context problem, and why pasting does not solve it
The usual workaround for writing a novel in a chat assistant is to paste the relevant chapter, plus a summary of the plot, plus a reminder of who the narrator is — every single session. It works, in the sense that you get an answer. It fails in the sense that you are doing the remembering, and the quality of the help is capped by how much you can be bothered to paste.
Muze Writer inverts that. Your premise, characters, voice and outline live in Story Core and are read automatically on every request, so the Muse already knows who your narrator is before you ask it anything.
A review that reads the entire manuscript
Ask a chat assistant to check your novel for continuity and you will be asked to paste it, in pieces, and it will answer about the pieces. The errors that matter in a long book are precisely the ones that do not live inside any single piece: a character whose description drifts, a subplot set up and never paid off, a timeline that stops holding in chapter 19.
Manuscript review reads the whole book in one pass and reports across it. Two speeds: minutes, or a deeper read that runs in the background. It costs credits in proportion to length, and the editor quotes the exact figure before anything is charged.
Rules it cannot break
Prompting a general assistant to avoid something works until it does not — a few thousand words later the metaphors are back. Muze Writer takes absolute constraints in Story Core, one per line: no metaphors, never start a chapter with weather, keep sentences under 25 words. They are injected last, into everything written for that book, including Rephrase.