According to Spotify's disclosure during its fourth-quarter earnings call this week, the company's top developers "haven't written a line of code since December." The stunning announcement was made by Spotify co-CEO Gustav Soderström, who also explained how the company is using artificial intelligence technology to speed up product development.
AI coding technology seems to have reached a turning point at Spotify. The music streaming giant revealed that its engineers are using an internal system called "Honk" to improve encoding speed and product development efficiency. The system uses generative AI technology, especially the Claude Code tool, to achieve remote real-time code deployment.
Soderström gave a concrete example on the call to show this technology in action: "A Spotify engineer on his morning commute can tell Claude to fix a bug or add a new feature to the iOS app through Slack on his phone. When Claude completes the work, the engineer will receive a push of the new version of the app on his phone's Slack, which can then be merged into production, all before they arrive in the office." Spotify believes the system "dramatically" speeds up coding and deployment.
In terms of product innovation, Spotify launched more than 50 new features and improvements to its streaming app throughout 2025. In recent weeks, the company has launched a number of new AI-driven features, including suggested playlists, page matching for audiobooks, and features such as "About This Song."
Soderström also highlighted Spotify's ability to build unique data sets, a competitive advantage that other large language models cannot easily replicate. He explained that unlike online resources such as Wikipedia, there often is no single factual answer to music-related questions. Taking fitness music as an example, people from different regions will give completely different answers: Americans generally prefer hip-hop music, but millions of people also like death metal; many Europeans choose electronic dance music as fitness music, while many Scandinavians love heavy metal.
"This is a data set that we're building on that no one else is really building on. It doesn't exist at this scale. And we see that every time we retrain the model, it keeps improving," Soderström noted. "We foresee that this is not the end of AI development, but just the beginning," the executive said.
During the earnings call, analysts also asked Spotify about its stance on AI-generated music. The company explains that it allows artists and labels to note how a song was produced in track metadata, while still policing spam on the platform.