Deep Cuts

Discover new music, build your record crates.

Start a dig

Start from records you love, or describe a vibe.

Up to three things you love: an album, a song or an artist

This week's dig

(example)

Taste profile

Your crates

How it works

  1. Start from records you love, or a vibe. An album, a song or an artist, or words like a place, a decade and a mood.
  2. It follows real links. The musicians and producers on your record, its label, band members and the records it samples, from MusicBrainz and Discogs. A vibe becomes Discogs searches by style, country and years.
  3. An AI chooses and explains. It picks the best five from those real records and writes why. It can't make records up: every pick is a real release.
  4. Rarity comes from collectors. Discogs shows how many people own and want each record, so every dig mixes an Essential, Digger favourites and Deep cuts. Dig deeper goes rarer.
Why can a dig take half a minute?

MusicBrainz, the open music database Deep Cuts reads, allows about one lookup a second. Records someone dug from recently come back much faster.

Is my data stored?

No accounts, no ads and no cookies. Visits are counted anonymously (Vercel Web Analytics) so I can see how the site is used. Your digs, crates and taste profile stay in this browser. What you search is sent to the AI to make the dig; Deep Cuts doesn’t save it.

Which AI is it?

An open model (gpt-oss) on Groq’s free tier. It only chooses among real records the site found, and writes the reasons.

Is it free?

Yes: free, with no ads and no sign-up.

Why Deep Cuts. I think taste can be acquired: you listen, you discover records, and slowly a taste forms. I built this to help me start somewhere on days I don't know what to play, and to dig deeper into artists I love. It's open source and built with AI as a coding partner.

How it learns. Each dig is based only on what you enter. Mark records Loved to save them to your crate; your reactions build your taste profile and make sure no record is suggested twice. Now spinning. Each visit plays a song from a great album: a hand-picked record from the shop's crate, or one your own digs found but didn't show you, so it slowly follows your taste. The reasons are written by AI, so check credits on Discogs before quoting them. Open source on GitHub. Album data from MusicBrainz. Covers from the Cover Art Archive. Data provided by Discogs. This site uses Discogs’ API but is not affiliated with, sponsored or endorsed by Discogs. “Discogs” is a trademark of Zink Media, LLC.

Found a bug, or have an idea? Write to xenerr@gmail.com.

Every record here was found through credits and collections shared by music lovers on MusicBrainz and Discogs.