For decades, the definition of a software engineer was simple: they wrote code. That definition is breaking down inside one of the world’s largest audio companies. Spotify recently admitted that some of its most productive developers have effectively stopped writing traditional code altogether, relying instead on a new workflow that turns plain English instructions into finished products before the morning commute is even over. This shift raises a difficult question about what technical work actually looks like in the near future.
Key Takeaways
- Spotify shipped more than 50 new features and changes to its app during 2025.
- Spotify uses an internal system called “Honk” powered by Claude Code for development.
- Some Spotify developers have not written a line of code since December.
During a recent earnings call, Spotify co-CEO Gustav Söderström stated that the company’s best developers have not written a single line of code since December. Instead, they are using AI tools to do the heavy lifting. The company has built an internal system called “Honk” to manage this process.
This system allows engineers to build and fix software much faster than before. Spotify credits this new workflow for helping it ship over 50 new features in 2025, including AI-generated playlists and better audiobook matching. The company views this as the start of a new way of working, not just a temporary experiment.
The big deal
This matters because it moves AI coding from a novelty to a core business operation. We often hear about AI writing small scripts or helping students with homework. This is different. A major public tech company is running its product development on AI-generated code. It suggests that the technology is now reliable enough to handle complex, real-world applications used by millions of people.
It also signals a change in the job market. The role of a developer is shifting from a writer to an editor. Engineers at Spotify are spending less time worrying about syntax errors and more time directing the architecture of the app. If a developer can do their job from a phone on a train, the barrier to building software is lower than it has ever been.
How it works
The core mechanism is a conversation between a human engineer and an AI agent.
Think of it like a head chef running a busy kitchen. In the past, the chef had to chop every onion and stir every sauce personally. Now, the chef simply shouts an order to a line of sous-chefs who do the manual labor instantly. The head chef tastes the final dish, approves it, and sends it out to the dining room.
At Spotify, the engineer acts as the head chef. They send a message on Slack telling the AI to fix a bug or add a feature. The system, powered by Claude Code, writes the actual computer code. It then builds a new version of the Spotify app and sends it back to the engineer’s phone. The engineer tests it, approves it, and merges it into the main product. This entire loop can happen while the engineer is commuting, without ever opening a laptop.
The catch
The article doesn’t say how much this system costs to run. AI models that are smart enough to write complex code are computationally expensive. While it saves human time, the bill for the computing power could be significant.
There is also a limit to what the AI can do on its own. Spotify points out that its advantage comes from its unique data. Generic AI models know facts, but they do not understand subjective human preferences. For example, an AI might not know that Scandinavians often prefer heavy metal for workouts, while Americans might choose hip-hop. Spotify has to feed its own proprietary data into the system to make it useful. Without that specific data, the AI is just a generic tool.
Additionally, the company has to actively police the platform for spam. As AI makes it easier to create music, the volume of low-quality content increases. Spotify allows artists to label AI-assisted tracks in the metadata, but it still has to filter out the junk.
What now?
Spotify plans to expand this system and retrain its models to get even better at understanding music preferences. The company sees this as a competitive advantage that other AI models cannot easily copy.
If you are a software developer, you should start treating prompt engineering and AI management as primary skills, not optional hobbies. The industry is moving toward a model where you manage code rather than write it. Watch to see if other major tech companies report similar drops in manual coding hours in their next quarterly reports.
