A radar for your idea.
What a sweep does, what the daily watch does, and how to read a score.
1 · The sweep
When you press sweep, the page converts your text into an anonymous numerical fingerprint on your device and compares it against every recent paper, right there in the tab. That's also the privacy story in one line: the comparison comes to your idea, your idea doesn't go to the comparison. (Details, if you want them.)
2 · The daily watch
Every day we index every new arXiv submission — all fields, thousands of papers — into the same kind of fingerprint. If you've armed the radar, each day's papers are compared against your idea's fingerprint. When one crosses your chosen similarity threshold, you get exactly one email: the paper, the score, and a calm action plan. No blip, no email.
3 · When the radar blips
Calibrated on live data: near-identical write-ups score around 100%. Same subfield with a different contribution sits near 40–55%. Genuine "two groups, one idea" collisions land at 65–85%. The default 60% threshold catches collisions while staying quiet — expect roughly one email a month per idea, not one a day.
Every alert ships the same checklist: read it today · map your delta · timestamp your own preprint · check the venue's scoop-protection window · consider emailing the authors — convergence starts more collaborations than feuds. A near-miss handled in 48 hours is an anecdote; discovered at review time, it's a rejection.
What we store if you arm alerts
Your email, a label you chose, your threshold, and the idea's fingerprint. Delete any of it, anytime, from the manage page.
More than scoop protection
A sweep is a map of the work nearest to yours — useful well before any collision:
The nearest-papers list is an instant related-work scan — the neighbors you should be reading and citing, ranked by closeness, not keywords.
Armed alerts surface papers that belong in your bibliography as they appear, so the reference you'd have kicked yourself for missing arrives in your inbox.
See how crowded your idea's neighborhood is before you commit months — a quiet area or a stampede, told in one score.
A near-twin is often a future co-author, not a rival. Convergent work is how some of the best collaborations start.
Same radar, more reasons to keep it armed. More sources and views are on the way — tell us what you'd use.
FAQ
- Is a 70% similarity score "I've been scooped"?
- No. It means "read this today." The radar finds semantic neighbors; only you can judge idea-identity. Most blips end with you adding one related-work citation and sleeping better.
- What exactly do you store?
- Email, a label you chose, a threshold, and your idea's anonymous fingerprint. Not your text. Delete anytime from the manage page.
- Which fields are covered?
- All of arXiv — every category, every day. More sources (bioRxiv first) are on the roadmap; Pro subscribers vote on what's next.
- Could the fingerprint be reversed into my idea?
- No — it's deliberately lossy. At most it hints at broad territory ("something about protein folding"), never your method, results, or words.
- Why don't you explain the machinery in more detail?
- The radar's internals are the part of the product we keep to ourselves. Everything that affects you — what runs where, what's stored, what an alert means — is documented on this page and the privacy page.
- What should I do if a real overlap appears?
- Don't panic-quit. Read it; map the delta; post your own preprint to timestamp priority; check the venue's scoop-protection policy (several journals allow ~6 months for in-progress work). The alert email includes this checklist.