Spotify Says Its Top Developers Haven’t Written Code Since December -Thanks to AI

Has AI-assisted coding finally hit a tipping point?

At least at Spotify, it might have.

During its latest earnings call, Spotify revealed something striking: according to co-CEO Gustav Söderström, some of the company’s best developers “have not written a single line of code since December.”

That doesn’t mean they’ve stopped building. Quite the opposite.

AI Is Now in the Driver’s Seat

Spotify says AI is dramatically accelerating how it develops and ships products. In 2025 alone, the company rolled out more than 50 new features and updates to its streaming app.

Recent launches include:

• AI-powered Prompted Playlists
• Page Match for audiobooks
• About This Song

All three were introduced within just the past few weeks.

Behind the scenes, engineers are using an internal AI-powered system called “Honk” to boost coding speed and product velocity. The system relies heavily on generative AI — specifically Claude Code — to write, fix, and deploy code.

Söderström shared a particularly futuristic example: a Spotify engineer commuting to work can message Claude from Slack on their phone to fix a bug or add a feature to the iOS app. Claude generates the changes, pushes a new version back via Slack, and the engineer can approve and merge it into production — all before arriving at the office.

In other words, coding can now happen from the passenger seat of a train.

Spotify says this workflow has “tremendously” sped up development and deployment. And the company believes this is just the beginning of what AI can unlock.

Building a Dataset No One Else Has

Beyond coding, Spotify also highlighted a longer-term AI advantage: its unique data.

Unlike general knowledge sources such as Wikipedia, music preferences don’t have single factual answers. What counts as “workout music,” for example, varies wildly by person and geography.

In the U.S., hip-hop may dominate gym playlists though plenty prefer death metal. Across Europe, many gravitate toward EDM, while Scandinavians often favor heavy metal.

Spotify argues that this kind of taste-driven, behavioral dataset is something few others can replicate at scale. And as it retrains its models, that dataset keeps improving.

What About AI-Generated Music?

Analysts also pressed the company on AI-generated tracks.

Spotify’s approach so far is to allow artists and labels to disclose in metadata how a song was created. At the same time, the company says it continues to police the platform for spam and low-quality uploads.