Has AI coding reached a tipping point? At least that seems to be the case for Spotify, which said on its fourth-quarter earnings call this week that its talented developers “haven’t written a single line of code since December.” This statement from Spotify co-CEO Gustav Söderström came alongside other comments about how the company is leveraging AI to accelerate development.
Notably, Spotify noted that it released over 50 new features and changes to its streaming app throughout 2025. And it recently rolled out even more features, including AI-powered prompt playlists, audiobook page matching, and About This Song, all of which were released within the past few weeks.
At Spotify, engineers use an internal system called “Honk” to improve coding and product speed, the company told analysts on a conference call. This system enables things like generative AI, especially remote real-time code deployment using Claude Code.
“As a concrete example, a Spotify engineer can ask Claude to fix a bug or add a new feature to an iOS app from Slack on his phone during his morning commute,” Söderström said. “Once Claude is done with that work, engineers can grab a new version of the app, push it to Slack on their phones, and merge it into production before they get to the office.”
Spotify credits the system with “significantly” speeding up coding and deployment.
“We predict that this is just the beginning, not the end, of AI development,” Söderström said.
The executive also touted Spotify’s ability to build unique datasets that other LLMs cannot commoditize, similar to other online resources such as Wikipedia. That’s because there aren’t always factual answers to music-related questions, he says.
tech crunch event
boston, massachusetts
|
June 23, 2026
For example, if you ask people what music is for working out, you might get different answers from different people based on geography. Americans tend to prefer hip-hop overall, but millions prefer death metal. Many Europeans go for EDM, while many Scandinavians like heavy metal.
“This is a dataset that we’re building right now that no one else is really building. It doesn’t exist at this scale, and we see it getting better every time we retrain the model,” Söderstrom said.
Analysts on the call also asked about Spotify’s approach to AI-generated music. The company explained that while it allows artists and labels to indicate how a song was created in a song’s metadata, it still polices spam on the platform.
