About

What this is

Arcurian is a way to go deep on anything you’re curious about. Type in a hobby, a craft, a specific object — and get a curated page revealing the corners of it a normal search doesn’t surface: the makers who don’t advertise, the specialist sellers, the people who’ve spent decades on one narrow question.

The problem we’re built around

Search engines are good at finding what’s popular. That’s exactly the problem: the ten links everyone already gets for “best fountain pen” are the ten links everyone always gets. The genuinely useful stuff — a specific model’s known failure point, the one small workshop that still repairs a discontinued mechanism, the collector who’s mapped an entire product line by hand — usually isn’t optimized to rank. It lives in comment threads, specialist communities, personal sites, and sellers that don’t show up on page one.

Large language models inherit the same bias. Ask one for recommendations and it tends to hand back the most-cited, most-written-about answers — which are the popular ones, not necessarily the best ones.

How we find it

We don’t just ask a model and publish what comes back. Finding the overlooked, genuinely good stuff takes real legwork:

• Following communities outward — starting from one relevant community and tracing its connections to the next.
• Reading discussion threads, not just top-level posts — the best specific recommendations usually show up in replies.
• Building source lists by hand, one hobby at a time, and deliberately weighting down the most obvious sources once the well-known ones are covered — the opposite of how most ranking works.
• A human review pass before anything is promoted to the homepage — a classifier flags candidates, but front-page visibility is curated, not automatic.

The synthesis — turning that research into a readable page — does use AI. The sourcing, which is the actual hard part, doesn’t start or end with a model call.

Who this is for

Built by people who are genuinely into a few specific things ourselves, for people who are too — not a one-and-done search, but somewhere that keeps surfacing the next layer.

Get in touch

Questions, corrections, or something we got wrong? Email [email protected].