A desk that knows
where to stop.
Put one research question and your sources on the desk. The Studio drafts your Introduction and Literature Review from them, and leaves the rest of the manuscript to you.
What does the desk actually draft?
The desk drafts two sections: an Introduction that frames your question, and a Literature Review built from the sources that survived screening. That is the whole promise, and it is printed on the screen the entire time you work.
It is not a limit we plan to lift later. A tool that writes your Methods and Results is writing your paper, and your reviewers are right to ask who did.
- Introduction and Literature Review, nothing beyond them.
- The scope band stays on screen, so nobody meets the limit halfway through a draft.
- The outline builder keeps both sections and refuses a conclusion, so the rule holds on the server too.
How is this different from a chat window?
A chat window will write any section you ask for and take the reference list on faith. The desk is built around the places where that goes wrong.
- Sections it will draftThe StudioIntroduction and Literature Review, then it stopsA general chat assistantWhatever you ask for
- Where the references come fromThe StudioYour bookshelf, your files, and a search you approvedA general chat assistantWhatever the model recalls
- Your approval before any provider is calledThe StudioIncludedA general chat assistantNot included
- Citation markers tied to a reference listThe StudioIncludedA general chat assistantNot included
- Download held back while claims are unverifiedThe StudioIncludedA general chat assistantNot included
- Restorable revisions of your own editsThe StudioIncludedA general chat assistantNot included
Where do the sources come from?
From you. Send items across from the bookshelf you built inside ScholarGate, or upload your own .bib and .ris files. They are read on the desk, and it tells you how many entries went in rather than claiming they all did.
The files you upload stay with the desk assistant. When you start a review, the pipeline runs its own literature search across open catalogs, and a desk reference carrying a DOI can join that source list once screening has run.
- Bookshelf items you sent to the desk.
- Your own .bib and .ris files, read on the desk.
- A desk reference with a DOI can be verified and added to a review's sources.
What is actually on the desk?
Three things, and none of them hidden. What you see on the desk is exactly what a review receives when you hand it over.
Who owns the words?
You do, at every gate. No provider is called until you approve the search plan. No prose is written until you approve the outline. Nothing is submitted anywhere, ever.
Once a draft exists you can rewrite it in place. Every save becomes a new version, earlier versions stay restorable, and regenerating over your own text asks first.
- You approve the search plan before any source is fetched.
- You approve the outline before a single paragraph is written.
- Your edits are versioned, and old versions can be brought back.
How does a draft leave the Studio?
As Word or Markdown, with the citation markers intact and the reference list assembled from the sources the review actually used.
Export is a gate, not a button. While the claim audit still has sentences it could not verify against their sources, the download stays closed and the flagged claims are shown to you instead.
- Word and Markdown, with citations and a reference list.
- Unverified claims keep the export closed until they are fixed.
- Drafting runs in English today, so the interface language does not change the manuscript.
Questions people actually ask
No. It drafts an Introduction and a Literature Review and stops. Methods, Results and Discussion are deliberately out of scope, and the server keeps them out even if a request asks for them.
BibTeX (.bib) and RIS (.ris) reference files, plus anything you send over from your ScholarGate bookshelf. The desk reports how many references were added and says so when the file held nothing it could read or the desk is already full.
No, and this surprises people, so it is written on the review form too. Uploaded files ground the desk assistant. A review runs its own literature search across open catalogs, and a desk reference with a DOI can be added to its sources after screening.
Yes, in place. Every save is a new revision, earlier revisions stay restorable, and if you ask to regenerate over text you edited yourself, the Studio warns you before it does.
Running a review needs one, because search, screening and drafting do real work with external providers on your behalf. The Review Studio is in early access and is not switched on for every account yet.
Put your question on the desk.
Bring one research question and the sources you already trust, and see how far two well grounded sections get you.
Open the desk