Spotify reports that thanks to AI, its best developers haven't written a line of code since December

Music streaming service

https://techcrunch.com/2026/02/12/spotify-says-its-best-developers-havent-written-a-line-of-code-since-december-thanks-to-ai/

On February 10, 2026, Spotify announced its financial results for the fourth quarter of 2025 (October to December). Revenue increased 13% year-on-year to $4.5 billion, with a gross profit margin of 33.1% and an operating profit of 701 million euros. The company's quarterly net increase in monthly active users reached a record high, meaning that at the time of writing, more than 751 million people worldwide were using Spotify monthly. The number of paid subscribers increased 10% year-on-year to 290 million.

During the company's fourth-quarter earnings call, co-CEO Gustav Söderström, speaking about Spotify's use of AI in software development, revealed that the company's engineers 'haven't written a single line of code since December.'

Spotify released over 50 new features and changes throughout 2025. In 2026, it also announced new features like 'Prompted Playlists,' an AI-powered playlist creation tool; 'Page Match,' which syncs physical books with audiobooks; and 'About This Song,' which lets you explore the story behind a song.

Spotify announces 'Page Match' feature that syncs physical books and audiobooks, and plans to start selling physical books as well - GIGAZINE

Spotify told analysts on a conference call that its engineers are using an internal system called 'Honk' to speed up coding and product development. Honk is a development tool that uses generative AI (specifically Claude Code) to enable remote, real-time code deployment.

'For example, Spotify engineers can ask Claude to fix bugs or add new features to the iOS app on their morning commute via Slack. When Claude is done, the engineer receives a new version of the app via Slack on their phone, which they can then merge into the production environment. And all of this is done before the engineer even arrives at the office,' Söderström said proudly.

Spotify credits Honk with helping it significantly accelerate its coding, and CEO Söderström said, 'We believe this is not the end, but just the beginning, when it comes to AI development.'

Spotify values large-scale language models (LLMs) for their ability to build unique datasets that can't be commoditized like online resources like Wikipedia. 'LLMs are especially useful because music-related questions don't always have factual answers,' Söderström said.

For example, if you ask, 'What music do you like to work out to?', the answer will likely vary from person to person. Americans tend to prefer hip hop overall, but many also prefer death metal. In Europe, many people prefer EDM while working out, while Scandinavian people tend to prefer heavy metal.

'This is a dataset that we're building, and nobody else is building anything like this. There's no other dataset of this scale. And every time we retrain our AI models, we see improvements in the dataset,' Söderstrom said.