The editorial argues Paramount is divesting the longest continuously-maintained record of human listening behavior in machine-readable form, not merely a music service. Twenty-four years of scrobbles, timestamps, and implicit-feedback signals constitute a data asset no streaming competitor can match in longitudinal depth combined with API openness.
The editorial positions the Last.fm-1K and Last.fm-360K datasets as the MNIST of music recommendation — not state-of-the-art, but the universal baseline that every implicit-feedback collaborative filtering paper has to beat. Their continued availability under independent ownership matters for academic and industrial ML research that has depended on them since 2010.
By submitting the support-forum announcement and driving it to 683 points with 185 comments, the submitter signaled that Last.fm's escape from CBS/Paramount after 19 years is significant news for developers. The unusual sustained attention for a service many assumed was moribund reflects a viewpoint that scrobbling's quiet endurance — and its potential rebirth as an independent entity — matters.
The editorial notes CBS paid roughly $280M in 2007 when social music graphs looked like the obvious future of streaming, but Spotify's 2008 global launch rendered that thesis quaint. The 19-year corporate ownership is framed as a strategic mismatch — Last.fm survived inside a media conglomerate that never figured out how to capitalize on its core asset.
Last.fm announced on its support forum that it is once again an independent company, severed from CBS Interactive / Paramount Global after a 19-year run as a corporate asset. The Hacker News thread crossed 680 points within hours — unusual sustained attention for a service most developers assume died sometime around the second Obama administration.
The facts are sparse and deliberately so. The post confirms a management buyout-style separation, retention of the existing engineering team, and a commitment to keep scrobbling free. There's no mention of new investors, no funding figure, and no roadmap beyond "more of what you already use." CBS paid roughly $280M for Last.fm in 2007, at a moment when social music graphs looked like the obvious future of streaming. Spotify launched globally the following year and rendered that thesis quaint.
What Paramount is divesting is not a product — it's the longest continuously-maintained record of human listening behavior that exists in machine-readable form. Scrobbling began in 2002. Twenty-four years of timestamps, user IDs, track metadata, and implicit-feedback signals sit in that database. No competitor — not Spotify Wrapped, not Apple Music replay, not YouTube Music history — has both the longitudinal depth and the open-ish API surface to match it.
If you've taken a recommender systems course in the last fifteen years, you've touched Last.fm data. The Last.fm-1K and Last.fm-360K datasets, released by Òscar Celma in 2010, are still the default benchmark for implicit-feedback collaborative filtering. They show up in the original neural collaborative filtering paper, in the two-tower retrieval literature out of Google, in basically every "we built a better recommender" preprint on arXiv. The dataset is to music recommendation what MNIST is to vision: not state-of-the-art, but the universal baseline everyone has to beat.
That matters now because the data moat just changed hands. Under CBS, the scrobbling API was a sleepy cost center — kept alive because killing it would have generated bad press disproportionate to the savings. Under independent ownership, every byte of egress has to justify itself. The interesting question isn't whether Last.fm survives. It's whether the free, generous API survives.
Compare the trajectory of every other consumer data API that went independent or got acquired in the last decade. Twitter's free API ended in 2023 — the $42K/month enterprise tier killed a generation of academic NLP work overnight. Reddit's free API ended the same year, and we got the Apollo shutdown and a developer revolt that still hasn't healed. Imgur tightened terms in 2024. Genius lyrics deprecated their public endpoint. The default trajectory of every "free forever" developer API under new ownership is the same: API key required, then rate limits, then paid tiers, then sunset.
Community reaction on Hacker News is split. The top comment celebrates "the open web winning one" — a sentiment that feels nice and probably doesn't survive contact with a Series A pitch deck. The second-most-upvoted comment is more cynical: "They have one shot to monetize the data before someone trains a foundation model on the export." That's not paranoia. Spotify's Annoy library, the music similarity research out of Deezer, and the listening-history-as-language-model line of papers (Pandora, ByteDance) all suggest the same conclusion: aggregated listening data is now training data, and training data is now revenue.
The other angle worth naming: Last.fm has been quietly subsidizing a small ecosystem of indie clients — Pano Scrobbler on Android, Web Scrobbler in browsers, Maloja for self-hosters, the mpd-scrobbler daemons that live in Linux user-space. None of these have business models. All of them assume the upstream API will stay open and unauthenticated-write-friendly. If that assumption breaks, the longest-running open music graph fragments along with it.
If you're building anything that touches the Last.fm API today, treat this announcement as a deprecation warning and act accordingly. Three concrete moves:
Export your data now. Last.fm's GDPR export endpoint still works and dumps your full scrobble history as CSV. Independent ownership means a new privacy policy, new T&Cs, and a non-zero probability that bulk export becomes a paid feature. Run the export. Stash it in S3. You'll thank yourself in 2027.
Audit your scrobble-write integrations. If your app writes scrobbles (a media player, a Plex plugin, a homelab dashboard), check whether you're using a per-user OAuth token or a shared API key. Per-user tokens are durable across most policy changes. Shared keys get nuked the moment someone decides to enforce "one app, one key." Spotify did this in 2022 and broke half the unofficial clients overnight.
If you're shipping a recommendation system, lock in your evaluation dataset. The Last.fm-1K and Last.fm-360K datasets are still hosted at multiple academic mirrors, but the canonical link goes through Last.fm's research page. Mirror it locally. Document the SHA. Cite the snapshot date. Reproducibility shouldn't depend on a startup's S3 bill staying paid.
For music-app builders specifically, the strategic question is whether to start migrating off Last.fm scrobbling as the canonical source of truth. Self-hosted alternatives — Maloja, ListenBrainz from MetaBrainz — exist and are getting better. ListenBrainz in particular has an explicit "we will never charge for the API" commitment in its non-profit charter, which is more contractual protection than any independent startup can credibly offer.
The honest read is that independence is good news for Last.fm-as-a-product and ambiguous news for Last.fm-as-a-platform. The team that knows the codebase is staying. The cultural assets — the user reviews, the tag taxonomy, the artist pages that still outrank Wikipedia for niche metal subgenres — get a stewardship upgrade. But independence also resets the monetization clock to zero. Within 18 months, the new entity will either have a sustainable revenue line or it will be raising on the strength of its data. Both outcomes have implications for the free API, and neither outcome favors the indie developer who's been scrobbling silently since 2009. Watch the changelog. Mirror your data. Don't assume the next decade looks like the last one.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.