Spotify's Top Engineers Stopped Writing Any Code in December—Here's How AI Made It Possible

When Spotify’s co-CEO Gustav Söderström revealed during the company’s latest earnings call that its elite engineering team hasn’t written code manually since December, it marked a profound shift in how modern software development actually works. The music streaming giant didn’t just adopt AI as a helpful tool—it fundamentally rewired its entire development process around artificial intelligence, proving that the future of coding might not involve writing code at all.

This transformation has been nothing short of remarkable. Throughout 2025, Spotify shipped over 50 updates and new features, with recent releases including AI-driven Prompted Playlists, Page Match for audiobooks, and About This Song. What makes this pace possible? An internal AI platform called Honk that’s redefining what engineers actually do each day.

The Birth of Honk: AI Takes Over the Keyboard

At the heart of Spotify’s revolution sits Honk, an internal platform powered by generative AI technology—specifically Claude Code from Anthropic. This isn’t just a code suggestion tool; it’s a complete reimagining of the development workflow.

Söderström painted a vivid picture of how it works in practice. Picture a Spotify engineer heading to the office during their morning commute. Instead of waiting to sit at their desk, they open Slack on their phone and type an instruction to Claude: “Fix this iOS app bug” or “Add this feature to our platform.” The AI doesn’t just suggest code—it actually writes, tests, and deploys it. Within minutes, the engineer receives a notification via Slack with a fully prepared app version ready to merge into production. They can approve and ship features before they even arrive at the office.

This remote, real-time deployment capability has fundamentally compressed development timelines. What once took days of manual coding, testing, and debugging now happens in hours or even minutes.

When Engineers Stop Writing Code, What Do They Actually Do?

The shift raises an obvious question: if Spotify’s best coders aren’t writing code anymore, what are they doing? The answer reveals something important about the future of engineering roles. Rather than typing commands into IDEs, these engineers are becoming architects, reviewers, and decision-makers. They’re setting the high-level requirements, validating AI-generated solutions, and ensuring quality standards remain uncompromised.

The efficiency gains are undeniable. Spotify’s engineering teams can now iterate faster, test more variations, and deploy features at a velocity that would have seemed impossible in the pre-AI era. The role hasn’t disappeared—it’s evolved.

Spotify’s Secret Weapon: Proprietary Music Data at Scale

While other companies rely on standard datasets scraped from Wikipedia and public sources to train their language models, Spotify has built something far more valuable: a massive, proprietary dataset focused specifically on music. This isn’t just a competitive advantage—it’s becoming Spotify’s moat against larger AI competitors.

Music recommendations don’t follow universal rules. Ask a thousand people what’s best for workout music, and you’ll get a thousand different answers. Americans might say hip-hop, while others prefer death metal. Europeans gravitate toward EDM for exercise, but Scandinavians consistently choose heavy metal. Regional preferences, cultural nuances, and personal taste variations matter enormously.

Spotify’s dataset captures all of this nuance at a scale no other AI company possesses. As the platform processes millions of user interactions daily, its AI models become smarter about music context—something generic LLMs simply cannot replicate. “We are developing a dataset at a scale that no one else has, and it continues to improve as we retrain our models,” Söderström explained.

Managing the AI-Generated Music Question

As AI begins creating music itself, Spotify acknowledged during the earnings call that it’s navigating new territory. The company now allows artists and record labels to specify in a track’s metadata whether the music was AI-generated, offering transparency to listeners. Simultaneously, Spotify maintains active monitoring systems to prevent spam and low-quality AI-generated content from overwhelming the platform.

This balanced approach—enabling innovation while protecting the creator ecosystem—signals how Spotify plans to lead rather than simply react to AI-generated music trends.

What This Means for the Industry

Spotify’s engineers stopping manual coding isn’t a dystopian scenario where humans become obsolete. Instead, it represents a maturation of how technology companies can leverage AI to amplify human capability. The best engineers at Spotify aren’t being replaced—they’re being liberated from repetitive, mechanical tasks to focus on strategy, architecture, and quality assurance.

The question isn’t whether other tech companies will follow this path. They almost certainly will. The question is whether they’ll have the infrastructure, talent, and strategic vision to make it work at Spotify’s scale. For now, Spotify’s engineering team has become a case study in what’s possible when you stop asking “Can AI write code?” and start asking “How do we build around AI writing code?”