Spotify says top developers have stopped manual coding

Spotify co-CEO Gustav Söderström stated that the company’s best developers “have not written a single line of code since December.” This shift highlights how the music streaming company is changing its software engineering workflows through automation, moving away from manual coding for its most experienced engineers. Söderström framed this transition not as the final stage of the company’s technological evolution, but rather as the beginning of its journey in artificial intelligence development.

To achieve this, Spotify engineers are using an internal system called Honk, which integrates Claude Code, a generative AI tool. The system allows developers to deploy code and fix bugs in real time, even when they are away from their desks. Söderström illustrated this workflow by explaining that an engineer on their morning commute can use Slack on their mobile phone to instruct Claude to fix a bug or add a new feature to the iOS app. Once the tool completes the task, the engineer receives a new version of the app pushed directly to them via Slack on their phone. This allows them to review and merge the changes into production before they even arrive at the office.

This setup helped Spotify ship more than 50 new features and changes to its streaming app throughout 2025. Beyond accelerating software development and product velocity, the company is using these interactions to build a unique dataset. Söderström claims this dataset cannot be easily commoditized by other large language models in the way public online resources, such as Wikipedia, are. Because music preferences are highly subjective, there is rarely a single factual answer to music-related queries, making the data highly specialized.

The dataset focuses on these subjective queries where preferences vary significantly by geography and taste. For example, workout music preferences differ across regions. While Americans generally prefer hip-hop, millions of others prefer death metal. Similarly, while many Europeans choose electronic dance music for workouts, many Scandinavians prefer heavy metal. Spotify is building this dataset at scale, retraining its models to capture these regional nuances. Söderström noted that no other company is currently building a dataset of this scale, and its quality improves every time they retrain their models.

Why it matters

Spotify is demonstrating a shift in software engineering where generative AI tools like Claude Code, integrated through internal systems like Honk, replace manual coding for its best developers. By capturing these developer and user interactions, the company is building a unique dataset that could serve as a defensible moat against commoditization by other AI models.