Do I need to create an account?
No. There are no accounts, no sign-up, no password. You paste your own API key and start — the only optional fields anywhere are your ORCID iD and institution (see Privacy), and those are just text boxes, not a login.
What happens to my API key?
It’s held only in this browser tab’s memory for the current session, sent directly from your browser to your chosen provider on each request, and never touches any server we operate. Close the tab and it’s gone.
Is my research actually private?
Yes — this is a static HTML file with no backend. Your provocation, framework, and full transcript are transmitted directly from your browser to Anthropic or OpenRouter (whichever you picked) to generate replies. Confident-tier citations are automatically checked against CrossRef and OpenAlex as they appear, and at the Literature Debate stage your provocation and domain are automatically sent to Semantic Scholar, PubMed, Europe PMC, and DataCite to check for existing work (this one has an off switch, in the tool's Data & Privacy tab). ROR and ORCID are only contacted for those specific optional features. We never receive any of it.
What’s the ORCID field for?
Purely optional attribution. If you enter your ORCID iD, your profile’s public name and affiliation (if visible) are looked up automatically and, along with the iD, added as one line to records you export — nothing else changes, nothing is verified, nothing is sent anywhere except in a file you generate and choose to share.
How do I know the sparring partner’s citations are real?
Every citation carries a tier tag (Confident, Approximate, or Logical inference). Confident-tier citations are automatically checked against CrossRef and OpenAlex the moment they appear — no button to remember — with the result shown right under that message; the export menu also has a full-session summary. A "found" result confirms a real paper by that author and year exists, not that the specific claim is accurate. Always verify independently before academic use — AI-generated citations can still be wrong.
Can I jump straight to a later stage?
Yes — if you already have a framework or draft, choose "Jump to Stress-Test" or "Jump to Peer Review" on the starting screen and paste it in. This skips the earlier stages; the sparring partner engages with what you pasted directly.
What does the SHA-256 hash on my export prove?
That the exact file hasn’t been altered since you exported it — paste it into "Verify a record" any time and it’ll flag any change, even one character. It doesn’t prove the content is factually correct, only that it’s unmodified.
Does this tool store anything about me?
Only in your own browser’s localStorage — your current session, saved projects, and an optional ORCID iD/institution if you provide them. Clearing your browser data deletes all of it. Nothing is stored on any server, because there isn’t one.
What’s the "AI contribution" tracker?
A simple count of how many turns were yours versus the sparring partner’s, shown as a running split. It’s informational only — a reminder that the debate is a record of your intellectual work, not something to optimize.
What is the CRediT contribution statement?
A standard-format disclosure (credit.niso.org taxonomy) mapping which parts of the work were led by you versus the AI sparring partner, generated from the stages your debate actually reached. Increasingly expected by journals for AI-assisted research.
Can I get a permanent, citable copy of my debate record?
Yes — "Deposit to Zenodo" in the export menu uploads and publishes your record to your own Zenodo account (you’ll need a free account and a personal access token), which gives it a permanent DOI.