Dig
Wash a night of arXiv and keep the few stones worth a careful look. Everything scores in this tab — nothing about your topics leaves it. Small print.
A claim is one setup — scout window, touchstones, core samples, weights. Edits save themselves as you make them; Save as… keeps the current setup under a name you can come back to. Export… downloads it as a file; Import… loads one someone sent you.
Relay — advanced, leave blank
Browsers refuse to read a response from a site that does not opt in, and arXiv does not opt in. So the arXiv request is sent for you by arxave's relay. It carries a category list and public paper IDs — never your topics, your library, or a key.
Where things go. Touchstones, core samples, and weights stay in this tab. With in-browser embeddings nothing is billed and nothing is sent; only the arXiv fetch leaves, through the relay above.
What is real today. Scouting arXiv, embedding abstracts in your browser (staged, with the seam map as intermediate output), and ranking on touchstone similarity with live re-blending. The rush (Scirate) is parked behind a Cloudflare challenge and stays inactive until a path exists.