Turn a focused question and your source material into a structured first draft. The short path below is enough for a demo; specialist evidence and review tools remain available when you need them.
This page opens as a new project. Your work is still saved locally as you type, but an earlier output is shown only when you choose to restore it.
A saved project is available in this browser.
Give the paper one clear topic and, if possible, one answerable research question.
A draft can start without files, but supplied sources make provenance and verification stronger.
Upload article notes, abstracts, source exports, supervisor comments, or draft files. TXT, MD, CSV, JSON, HTML, RTF, DOCX, and legacy DOC are handled in the browser where possible.
Files are attached only to the current generation request and are not saved as uploaded library files. Reattach them after restoring a project. Do not attach confidential material unless you intend the selected AI route to process it.
These three choices control the document format, depth, and AI route.
OpenAI, Gemini, xAI, and Quantum lanes require the appropriate account route; ZeroThink applies its existing fallback rules when a lane is unavailable.
Generate article prose now. Planning and specialist evidence tools are in the panels below.
The draft remains unverified until a researcher checks its sources and claims.
Plan a reproducible search and label what is verified, supplied, or still only a lead.
Map the evidence needed, its provenance, test options, and the paper structure.
Create the alternative survey format, critique the current output, plan revisions, or stress-test its strongest claims.
Copy the current output, download it, preserve the project, or start again.
Ready. This is a new project; no saved output has been restored and no evidence claims have been generated.
Attach two to five seed articles, abstracts, BibTeX entries, or notes. Run Plan the Literature Search, Decide What Evidence Is Needed, Evidence Graph, List and Check Sources, Build Paper Structure, Create Full Paper Draft, Critique the Draft, and Plan Improvements. Treat every new source as unverified until a DOI, URL, uploaded text, or database result confirms it.
The source ledger is not meant to present AI-invented papers as real sources. Each row must show where it came from: an uploaded file, researcher notes, a retrieved search result, supplied full text, a candidate lead, or an item that still needs verification.
Text supplied by the researcher through the upload box or notes field. The system can quote or summarize it, but the researcher still checks the original file before submission.
Live web or Serper search context when enabled and available. This is a discovery result unless the output clearly states full-text access was provided.
A model-suggested title, author, query, or likely paper to search for. Candidate leads are not treated as verified citations until a DOI, URL, database result, or supplied file confirms them.
Anything missing a reliable source trail must be marked before drafting claims around it.
Every source, claim, and synthesis step is tied back to a ledger where possible.
The researcher can steer, correct, validate, and approve before final synthesis.
The Nottingham and Paper Creator lanes can build a claim/evidence/provenance graph, then propose constrained optimisation tests for evidence-path selection, clustering, routing, prioritisation, and auditability. Quantum or simulator methods must be compared with classical baselines and must not be described as understanding documents.
The workflow starts with literature and claims, not one-shot generation.
OpenAI, Gemini, xAI, Groq, NVIDIA, and QuantumZero can be selected for different research passes.
Arguments, counterarguments, gaps, and reviewer-style critique happen before the final draft.
A literature-research plugin, repository, invitation, or team project can supply sources and workflows, but it remains separate from the ZeroThink agent until an integration is deliberately implemented and verified.
This workspace does not claim university endorsement, institutional adoption, contributor availability, repository access, or formal project membership. Researchers remain responsible for source access, authorship, disclosure, ethics approval, and submission decisions.
Human verification is required. Check every citation against the original source, confirm quotation wording and page numbers, review statistical claims, disclose AI assistance where required, and follow the target venue's authorship and research-integrity rules.