A structured coach, not a chatbot

The four stages

1. Puzzle

   Surface the real puzzle behind your idea — what the field can't yet explain — pitched at the right altitude.

2. Research question

   Turn the puzzle into a specific, answerable question, with the theoretical tension it resolves made explicit.

3. Testing

   Pressure-test it: is the tension real, or is this gap-spotting? Does the contribution redraw the field's map, or just add to the pile?

4. Summary

   A clean write-up of your sharpened question, its contribution, the coach's assessment, and your reading list — yours to keep.

What a strong question needs

As you work, a live scorecard rates your idea on the criteria editors and reviewers actually use:

What happens to your data

Your ideas stay yours. In short:

• Your research ideas stay yours and stay confidential: nothing you share is ever shared outside the research team or published — results are reported only as anonymized, aggregated patterns.
• Participation is voluntary; you can stop at any time by closing the tab.
• Please do not paste or upload confidential or identifying information.
• Your conversation is processed by Anthropic's API to run the coach. Anthropic states that API data is not used to train its models and is deleted within 30 days (kept longer only if flagged for a safety review) — this is Anthropic's stated policy, which we relay but cannot independently guarantee. The literature search sends only short search terms to OpenAlex — never your full text.
• Data is stored in the EU and used only for research on the tool.
• We store a random identifier in your browser so we can recognise if you return to develop additional research questions. It contains no personal information, is used only to analyse repeat use of the tool, and you can remove it any time by clearing your browser data.

What it can — and can't — do

The journals it checks

The literature search runs against a fixed whitelist of peer-reviewed journals — no preprints, no open web. That's a deliberate guardrail against hallucination: rather than trusting the AI to recall references on its own, where fabricated citations creep in, it only draws on real, verified journals. It searches the whole list every time, using the field you give it to keep your own discipline covered and prioritized while still surfacing relevant work from neighboring fields. It can't read full papers, though — most are paywalled, so it works only from titles, abstracts, and metadata; its picture of the literature is inevitably partial and can be wrong, so treat it as a starting point and confirm against the sources yourself. The groups below are just for browsing. Missing a flagship journal in your field? Tell us.