Has AI coding reached a tipping level? That appears to be the case for Spotify not less than, which shared this week throughout its fourth-quarter earnings that one of the best builders on the firm “haven’t written a single line of code since December.” That assertion, from Spotify co-CEO Gustav Söderström, got here alongside different feedback about how the corporate is utilizing AI to speed up improvement.
Of be aware, Spotify identified it shipped greater than 50 new options and modifications to its streaming app all through 2025. And, most lately, has rolled out extra options, like AI-powered Prompted Playlists, Web page Match for audiobooks, and About This Music, which all launched inside the previous few weeks.
At Spotify, engineers are utilizing an inside system known as “Honk” to hurry up coding and product velocity, the corporate informed analysts on the decision. This technique permits for issues like distant, real-time code deployment utilizing generative AI, and particularly Claude Code.
“As a concrete instance, an engineer at Spotify on their morning commute from Slack on their cellphone can inform Claude to repair a bug or add a brand new function to the iOS app,” Söderström stated. “And as soon as Claude finishes that work, the engineer then will get a brand new model of the app, pushed to them on Slack on their telephone, in order that he can then merge it to manufacturing, all earlier than they even arrive on the workplace.”
Spotify credited the system in serving to to hurry up coding and deployment “tremendously.”
“We foresee this not being the tip of the road by way of AI improvement, just the start,” Söderström stated.
The exec additionally touted Spotify’s potential to construct a novel dataset that different LLMs couldn’t commoditize, the best way they might different on-line sources, like Wikipedia. That’s as a result of there’s not at all times a factual reply for music-related questions, he stated.
Techcrunch occasion
Boston, MA
|
June 23, 2026
As an example, should you requested what exercise music is, you’d get totally different solutions from totally different individuals, generally primarily based on their geography. People are likely to want hip-hop general, although hundreds of thousands want demise metallic. And whereas quite a lot of Europeans would work out to EDM, many Scandinavians like heavy metallic.
“It is a dataset that we’re constructing proper now that nobody else is basically constructing. It doesn’t exist at this scale. And we see it enhancing each time we retrain our fashions,” Söderström famous.
Analysts on the decision additionally requested about Spotify’s method to AI-generated music. The corporate defined that it’s permitting artists and labels to point out in a observe’s metadata how the music was made, however remains to be policing the platform for spam.
