One of the most delicious tech acronyms is MVP. It stands for “minimum viable product” — the quality line below which a product cannot go and still be shippable. (It’s fun to extend the idea to other fields. Minimum viable pop star: Andrew Ridgeley of Wham!; minimum viable NBA player: Brian Scalabrine.)

It’s a useful concept when you encounter something that misses that bare minimum, like the new ChatGPT integration with Spotify, which launched on Monday. It’s a combination that somehow makes both products worse. It’s so buggy it made me nostalgic for an era when tech companies felt obliged to release things that actually worked. It’s so broken, it felt like I should’ve been getting paid to use it as an alpha tester.

According to the announcement, users can now just mention Spotify in a ChatGPT prompt to connect the two apps. From there, they’re supposedly able to ask for things like “your favorite K-Pop star” or “a playlist featuring all the Latin artists in your heavy rotation.” Spotify Premium subscribers, the pitch goes, will see their “more elaborate prompts” turned into a “fresh and fully personalized selection of tracks.”

None of the above actually happened for me. When I asked to connect, ChatGPT didn’t just fail; it made fun of me for asking. 

It was a harbinger of things to come. One likely culprit is the Model Context Protocol — the system that lets large language models talk to other apps. When MCP works, it feels like magic: AI can actually do things, like send emails or build simple tools on command. When it doesn’t work, one hand has no idea what the other is doing.

During my test, ChatGPT went into full customer-support hallucination mode more than once, inventing fixes like a nonexistent “Plug-Ins” or “Labs” section to visit. Spotify wasn’t much better, often replying in its widget that “Spotify couldn’t create your playlist. Try asking for something else.”

Worst of all, the two services often couldn’t even agree on what was happening. Several times, Spotify threw up error messages while ChatGPT celebrated imaginary wins.

When Spotify did manage to serve up a playlist, it looked identical to what you’d get from its existing AI Playlist feature, which has been around for over a year (a lifetime in AI). AI Playlist is a decent tool for music discovery once you learn its quirks, though it often ignores parts of a prompt, like a request for a certain time window. If you want to try it, you’re better off avoiding ChatGPT and going straight to Spotify. Or just ask ChatGPT for music or podcast recommendations on its own — it does fine.

I’m guessing everyone involved knows this collab is undercooked and wants to fix it fast. The announcement even hinted as much: “It’s early days, so while we might not be able to deliver on every request just yet, we’ll continue to build, refine, and improve the experience over the coming weeks and months.”

But by midweek, the partnership itself seemed to be fraying, right there in the chat. On day three, Spotify failed a podcast request, and ChatGPT simply took over, offering its own recommendation.

It felt like ChatGPT was quietly giving up on its partner. As Spotify kept failing, it shifted into full-on snark.

By this time, my poor little brain was done as well.

How is this buddy movie going to end? Will this collab ever make it past the MVP line? Watch this space, and as always, send me thoughts and feedback to [email protected]

I bumped into “If You Know Me (demo)” through a random guitarist on Instagram who was trying to play its haunted main riff, and found myself digging through the comments to find the original: a pretty darned beautiful two minutes from Hudson Freeman, a Brooklyn-based, Texas-born son of Evangelical missionaries. Freeman recorded “If You Know Me” in a field in Indiana — if you listen closely, you can hear crickets in the background, a heavy vehicle roll past in the distance, and echoes of the late, grumpy fingerpicker John Fahey, who might be smiling down from somewhere, despite his better instincts.

Lots of ‘80s and ‘90s vibes in this one; let me know your best time!