Who it is for
Listeners who think in eras — who want to walk through a decade by year and genre — and then play the track on whichever service they already use.
How it works
It runs entirely in your browser. One engine drives the hub and the five era pages with no framework; each era’s tracks ship as a data file. The hub itself loads no track data, and long lists render in pages of 100 so a phone’s main thread does not lock up.
The shuffle is the careful part. It draws from crypto.getRandomValues, discards values that would bias the pick (rejection sampling), and deals from a Fisher-Yates bag, so no track repeats until the whole filtered set has played. The code’s own comment works through why naive random draws repeat about one time in ten over a hundred presses on a thousand tracks.
Links are built from artist and title as the page renders. A direct YouTube video is used only where its id survived an audit — the 70s file alone had 153 duplicate ids and 116 plainly generated ones — and every other track gets a search link that always lands on the song. Search, genre and sort live in the address fragment, so a filtered view such as #q=acid&g=Techno can be bookmarked.
What we tested
All times are UTC on 27 September 2026, measured against a copy of labs.llc build 537 served from our own machine, or against the public source the page itself calls.
- Hub, two era pages and the engineHub 200, 24,019 bytes; 60s page 100,655 bytes; techno page 87,490 bytes; songs.js 45,401 bytes.
- The catalogue, parsed60s 961 tracks (1955–1973); 70s 1,021 (1968–1988); 80s 974 (1976–1992); 90s 1,105 (1985–2002); techno 957 (1975–2019).
- Direct video links162 of 5,018 kept an audited id: 60s 37, 70s 120, 80s 5, 90s and techno none.
Six 90s rows are flagged “circa” where the year is uncertain. It is a small thing, but the catalogue says so instead of guessing.
How it compares
We checked each alternative’s own pages on 27 September 2026 and describe only what we saw there. Each gets the pick below when it does the job better.
Radiooooo
checked 27 Sep 2026A “musical time machine”: pick a country and a decade and it plays a human-curated selection, from the 1900s to today, on a world map, with a Taxi mode across several destinations, themed Islands, likes and crossfade. There is a free tier and in-app subscriptions. We checked it through its App Store listing, as the web app gave our fetch no readable content.
Better at
- It plays the music, continuously, like a radio.
- Countries worldwide and decades back to the 1900s.
- Human curation, moods and a multi-destination Taxi mode.
Where Song goes further
- The whole catalogue as a searchable, sortable list with year, genre and tag for every track.
- Each track handed to the service you already use — five of them.
- Free with no account, a filtered view you can share as a link, and a shuffle that will not repeat early.
Every Noise at Once
checked 27 Sep 2026Glenn McDonald’s interactive map of Spotify’s algorithmically generated genres — 6,291 named genres and about a million artists by December 2023 — with 30-second samples. It has been frozen since McDonald was laid off by Spotify that December. Its site refused our fetch, so we checked it through its Wikipedia article (linked).
Better at
- Vastly more genres and artists, laid out as a map you explore by ear.
- Audio samples in the page.
Where Song goes further
- Organised by decade and year, for listeners who think in eras rather than micro-genres.
- Five services per track instead of one, and direct videos only where the id was audited.
Where it falls short
- It plays nothing itself. Every track is a link out, and for 4,856 of 5,018 tracks the link is a search results page, not the recording.
- The catalogue is fixed at build time: no adding, rating or saving tracks beyond the fragment link.
- Five Anglo-American eras plus techno — nothing before 1955, and no view by country.
- Its portfolio blurb says “no tracking”, but the hub and era pages load Google Analytics 4 on labs.llc.
Which one to pick
| If you need… | Pick |
|---|---|
| Browse an era by year and genre, then play it on your own service | Songours |
| A shuffle that never repeats until the list is done | Songours |
| Press play and let a country and decade run | Radiooooo |
| Explore thousands of micro-genres by ear | Every Noise at Once |
The fit
Best for
- Finding a decade’s tracks to play on the service you already use
- A shuffle that deals every track once before any repeats
- Sharing a filtered list by link
Not for
- Listening without leaving the page
- Music from before 1955 or beyond the Anglo-American charts
- Anyone who reads “no tracking” literally — the pages load Google Analytics
Pick Song if you want to walk an era’s catalogue by year and genre, then play it wherever you already listen.
Sources for this review
- song/assets/js/songs.js in build 537 — header lines 1–23, shuffle at 25–95 and 571, links at 97–135, paging and fragment at 469–473, hub note at 899–900; song/assets/data/*.js; song/index.html lines 26–34
- Radiooooo’s App Store listing and the Every Noise at Once Wikipedia article, fetched on 27 September 2026 at about 20:04 UTC