Spotify is pushing the boundaries of what AI can do in software creation. During their recent earnings discussion for the last quarter, co-CEO Gustav Söderström revealed something striking: the company’s top programmers have not typed out any code themselves since late last year.
He mentioned this while explaining various ways the streaming giant is leveraging artificial intelligence to make building products much faster.
Last year alone, the platform introduced over fifty updates and fresh elements to its music and podcast app. Lately, several exciting additions have appeared, such as playlists created through simple user instructions powered by AI, a tool that connects audiobook pages smartly, and details explaining the background of individual tracks; all of these went live in just the past couple of weeks.
The team relies on their own custom setup named Honk to boost how quickly they produce and launch improvements.
This tool brings in advanced generative AI capabilities, including direct use of Claude for coding tasks, enabling instant updates and deployments from anywhere.
To show how practical it has become, Söderström described a typical morning routine: a developer, while traveling to work, opens Slack right on their smartphone and instructs Claude to resolve an issue or build something new for the mobile iOS version.
Moments later, the finished update arrives as a fresh app build straight through Slack. The engineer simply approves and sends it live to everyone, often finishing before stepping into the building.
This whole approach has dramatically increased the pace of writing code and getting changes out to users, according to the company.
Looking ahead, Söderström views the current stage as only the starting phase for deeper AI involvement in their engineering work.
He also highlighted a major advantage Spotify holds: they’re gathering a special collection of information tied to music tastes that generic AI models can’t easily copy or pull from public sources like encyclopedias. Music preferences don’t always have one clear right answer, he pointed out.
Take the question of ideal songs for exercising, responses vary widely depending on who you ask and even where they live. In the U.S., hip-hop often tops the list for many, yet plenty of people reach for intense death metal instead.
Across Europe, electronic dance music draws crowds to workouts, but in places like Scandinavia, heavy metal frequently wins out.
Söderström emphasized that this growing collection of real user music insights is something unique at this level, unmatched by others right now. Every time they update and retrain their systems, the quality gets noticeably better.
During the same session, questions came up about handling tracks made entirely by AI. Spotify lets creators and music companies tag songs in the background info to show if artificial intelligence played a role in production.
At the same time, the service keeps strict checks in place to block low-quality or spammy uploads from flooding the catalog.
Other Stories You May Like