Spotify’s Best Developers Haven’t Written Code by Hand Since December

Spotify just dropped one of the most significant signals yet about where AI is taking the workforce.

During its Q4 earnings call, co-CEO Gustav Soderstrom told analysts that Spotify’s best developers “have not written a single line of code since December.” Not AI-assisted coding. Not code with AI suggestions. Zero hand-written code for months, from the company’s top engineers.

The results: record gross margin, shares soaring, 751 million monthly active users, individual developer productivity up 30%, and more than 50 new features shipped throughout 2025. Roadmap items planned for two years out are being worked on this year.

This isn’t a startup experimenting with AI tools. This is a major tech company with hundreds of millions of users running its development operation through AI. And it’s working.

How HONK Changed Everything

The system behind the transformation is an internal platform Spotify built called HONK. It layers on top of Anthropic’s Claude Code and integrates directly into Slack, allowing engineers to interact with AI from their phones.

The workflow is striking: a Spotify engineer on their morning commute opens Slack on their phone, tells Claude to fix a bug or add a new feature to the iOS app, and receives a new version of the app pushed back to Slack before they arrive at the office. Bug fixes, feature additions, and code deployment, all from a phone, all without writing a single line of code by hand.

James Perkins puts it in practical terms: “You can write a prompt that says ‘Go test this segment of the code’ or ‘I found a bug with page load time. Can you go address this?’ You don’t have to give any more detail than that. And if you’re using Claude Code or Codex, it’ll go find the page you’re referencing, review the code on its own, find the bug, and fix it, and then write you back and say ‘it’s done, and I found these two other things.’”

What Engineers Actually Do Now

The shift changes the fundamental nature of the developer role. Engineers at Spotify are no longer writing syntax. They’re reviewing AI-generated output, making architectural decisions, and directing the system toward the right outcomes. Senior engineers act as reviewers, architects, and orchestrators. The role has shifted from code writer to AI supervisor.

James frames the broader trend: “Brilliant people aren’t going to stop being brilliant. They’re just going to do more. If you’re a great engineer, you’re able to code much more now with AI tools.”

He’s also seeing this pattern beyond Spotify: “Forward-thinking startups that are leaning into AI are definitely performing more, faster, better. I’m seeing CEOs that are really engaged in their business getting involved in coding again because you can use the tools to accelerate yourself so much.”

The Expansion: Video, Entertainment, and Compressed Roadmaps

The 30% per-developer productivity increase isn’t just about maintaining the status quo faster. Spotify is using the velocity to expand into entirely new territory.

James shared what he’s seeing from the company: “They’re going to start looking at doing video very aggressively on Spotify. Items from their roadmap two years into the future are being worked on this year. That’s how much faster they’re moving.”

The company is also exploring partnerships at the intersection of discovery and entertainment, incorporating new content formats and providers into the platform. Product backlogs that would have taken years to clear are being addressed in months.

Where Humans Still Matter

James is careful to note that significant gaps remain. UI/UX design is a prime example: “You can plug Figma into Claude or Codex and have it do some of that for you, but there’s still a need for tons of human intervention there. There’s no one de facto standard that works for everybody.”

He points to the concept of “AI slop,” where AI-generated interfaces default to the same generic look and feel. Subjective decisions about user experience, where multiple valid approaches exist, still need human judgment and taste.

His advice for developers: find where your passion and skill set intersect with gaps in what AI can do. Learn how to use AI for your specific interest areas. If you love testing, figure out how to use AI to build the best end-to-end tests. If you’re in UI/UX, understand where AI falls short and become the person who bridges that gap.

James also introduces an advanced concept: writing “skill files” for AI. These are essentially instruction sets that define a specific approach, methodology, or standard that the AI ingests and follows. This turns human expertise into a reusable asset that makes AI output better.

What Non-Developers Should Take Away

Sean extends the lesson beyond software engineering: “For my non-technical crowd, think of what this impact similarly might be for your world. You could be a project manager, product manager, customer service, whatever. Any of the stuff that you do today that theoretically can now or probably in the near future be automated or fully automated. Think about what your role looks like if you’re leveraging AI in that way.”

His framing: prepare for a future where humans don’t write code, and by extension, where AI handles the execution layer of most knowledge work. If that future doesn’t fully materialize, you’re still better prepared than before. If it does, you’re ahead of everyone who waited.

The Signal Is Clear

Spotify is the first major enterprise to publicly confirm a full transition from AI-assisted to AI-led development. Not a pilot. Not an experiment. A fundamental shift in how their engineering team operates, with measurable results: 30% productivity gains, 50+ features shipped, record financial performance, and compressed multi-year roadmaps.

The question for every knowledge worker isn’t whether this pattern spreads. It’s how quickly.

Sources

Check Your Career’s AI Risk

Our database covers every major profession with specific displacement scores, timeline estimates, and adaptation strategies.

• Take the career risk quiz → whataboutai.com/quiz
• Browse all jobs → whataboutai.com/jobs
• AI consulting for your business → whataboutai.com/business

Coaching: For personalized career guidance, visit whataboutai.com/coaching.