Spotify Developers Abandon Coding Thanks to AI Tool

Spotify has reached a significant milestone in its integration of generative AI, with co-CEO Gustav Söderström revealing that the company’s top developers have not manually written a single line of code since December. This shift is part of an aggressive strategy to automate software development and speed up the delivery of new features.

During the company’s fourth-quarter earnings call, executives highlighted that this increase in development velocity resulted in over 50 new features and system updates throughout 2025. Recent additions to the platform include AI-powered Prompted Playlists, Page Match for audiobooks, and the “About This Song” feature.

The ‘Honk’ System and Claude Code

The core of this transformation is an internal platform dubbed “Honk.” The tool integrates generative AI—specifically utilizing Claude Code—to facilitate remote, real-time code deployment.

Söderström provided a practical example of how the workflow functions:

• An engineer commuting to work can use Slack on their mobile device to instruct Claude to address a bug or implement a new feature for the iOS app.
• Once the AI completes the task, the engineer receives a new version of the app directly in Slack.
• The developer can then merge the update into production before even stepping into the office.

According to the leadership team, this system has “tremendously” improved the pace of deployment. Söderström characterized this current state as merely the beginning of the company’s AI-driven evolution.

Building a Proprietary Dataset

Beyond coding, Spotify is focusing on creating a unique dataset that distinguishes its AI capabilities from competitors. Söderström argued that while standard Large Language Models (LLMs) rely on public data like Wikipedia, Spotify’s data is distinct because music preferences lack a single “factual” answer.

For example, the definition of “workout music” varies significantly based on geography and personal taste:

• American listeners often gravitate toward hip-hop.
• European audiences frequently opt for EDM.
• Scandinavian users show a marked preference for heavy metal.

“This is a dataset that we are building right now that no one else is really building. It does not exist at this scale,” Söderström explained, noting that the data improves with every model retraining.

Regarding the broader implications of AI-generated music, Spotify confirmed it is allowing labels and artists to disclose the use of AI in track metadata. Simultaneously, the company continues to actively monitor and police its platform to mitigate spam and maintain content quality.