At Spotify, Senior Engineers No Longer Write Code

Spotify's most senior developers stopped writing code two months ago. They haven't stopped working. They just don't type code anymore.

"When I speak to my most senior engineers, the best developers we have, they actually say that they have not written a single line of code since December," co-CEO Gustav Söderström said during the company's February 10 earnings call. "They actually only generate code and supervise it."

The shift happened fast. December marked what Söderström called a "singular event" in AI productivity. Anthropic released Claude Opus 4.5 with its Claude Code tool that month, and according to Söderström, things "just started working."

Now, Spotify runs an internal system called Honk that lets engineers direct software development through Slack messages. An engineer can tell Claude to fix a bug or add a feature to the iOS app from their phone on the morning commute. Claude writes the code, pushes a new version of the app back to Slack for review, and the engineer can merge it to production before getting to the office.

"It is a big change. It is real, and it is happening fast," Söderström said.

From typing code to directing AI

The change goes beyond speed. It redefines what software engineers do. Instead of writing functions and debugging syntax, Spotify's developers now set requirements, review AI-generated code, and make architectural decisions. The actual coding work happens automatically.

Söderström said the shift will require "a lot of change" at tech companies that want to stay competitive. Engineering practices, product workflows, and design processes will all need to adapt.

The challenge is that the technology is still evolving.

"The tricky thing right now is that if this was the end of the change, you could say this is what happened. Now let us retool for this," he explained. "The tricky thing is that we are in the middle of the change, so you also have to be very agile."

The payoff shows in the numbers. Spotify shipped more than 50 new features in 2025, including Prompted Playlists, Page Match for audiobooks, and About This Song. Most launched in the past few weeks alone. The company credits Honk with making that velocity possible.

Building data AI models can't copy

Speed isn't the only advantage Spotify sees from AI. The company is building a proprietary dataset that general AI models can't replicate: the connection between natural language and subjective music taste across different cultures.

"There is no factual answer to what is workout music," Söderström noted. "For an American, it is usually hip hop. For a European, it is usually EDM. For many Scandinavians, it is something like heavy metal or even death metal."

Spotify collects this data from hundreds of millions of users in different markets who constantly tell the platform what songs mean to them in specific contexts. It's not something AI companies can scrape from public sources like Wikipedia.

"This is a dataset that we are building right now that no one else is really building. It does not exist at this scale," Söderström said. "And we see it improving every time we retrain our models."

That dataset powers features like AI DJ, which has been used by about 90 million subscribers and generated over 4 billion hours of listening time. It also creates a competitive barrier against both traditional streaming rivals and the AI companies whose tools Spotify uses for development.

Spotify reported Q4 revenue of $4.5 billion, up 13% year-over-year. Operating income hit €701 million. Monthly active users reached 751 million, with 290 million paid subscribers. Free cash flow for the quarter was $834 million.

The company raised its U.S. Premium subscription price to $12.99 in January 2026, the third increase in three years. The move came weeks before revealing how much of its development now runs on AI rather than manual coding.

Söderström expects the transformation to continue. "Software companies will start producing enormously more amounts of software," he said.