AI Coding: The End of Manual Programming
How Spotify's AI system "Honk" replaced manual coding and shipped 50+ features in 2025.
Feb 14, 2026 - Written by Lorenzo Pellegrini
This image is part of Spotify's official brand assets, available from their press kit
Lorenzo Pellegrini
Feb 14, 2026
Spotify's Top Engineers Haven't Written Code Since December: AI's Bold Takeover
Spotify just revealed a game-changing shift in software development. Its most senior engineers have not touched a single line of code since December 2025, fully handing the reins to an internal AI system. This move signals AI's rapid evolution from assistant to primary coder in big tech.
The Announcement That Shook Tech
During Spotify's Q4 2025 earnings call on February 10, 2026, co-CEO Gustav Söderström shared stunning news. The company's top developers, those with the deepest expertise, stopped manual coding in December. They now generate code via AI and supervise its output. This transition followed the release of advanced AI models like Claude Opus 4.5.
Spotify reported massive growth alongside this change: 38 million new monthly active users, €701 million in operating income, and €4.5 billion in revenue, up 13% year-over-year. The company paid $11 billion to music rights holders in 2025, proving business thrived amid the AI pivot.
Meet Honk: Spotify's AI Coding Powerhouse
Honk is Spotify's custom AI platform, built on Anthropic's Claude Code. It handles everything from bug fixes to new features. Engineers interact via Slack from their phones, even during commutes. For instance, one can request an iOS app update, receive a ready-to-deploy build, and merge it to production before arriving at work.
This system enabled Spotify to launch over 50 features in 2025 alone. Highlights include AI-powered Prompted Playlists, Page Match for audiobooks, and About This Song. Development speed has surged, turning weeks-long tasks into hours.
What Do Engineers Do Now?
Roles have transformed dramatically. Senior engineers act as architects, strategists, and editors rather than coders. They define requirements, oversee AI output, and focus on high-level decisions like product architecture and complex problem-solving.
Söderström noted the bottleneck shifted from coding to human judgment. Engineers describe needs in natural language; AI executes. Human review ensures quality, blending speed with accountability. Spotify prepared for this by building an AI-native environment 18 months ago.
- Describe features or fixes via Slack.
- AI generates and tests code automatically.
- Review, approve, and deploy from mobile devices.
- Prioritize strategy over keystrokes.
Why This Matters for the Tech Industry
Spotify's success validates AI-dominant programming. Reliability comes from Claude Code's prowess in complex tasks, enhanced by Spotify's proprietary music data for personalized features. Traditional metrics like lines of code written matter less; velocity and impact dominate.
Concerns linger about code quality and error risks, but human oversight mitigates them. This model scales production while freeing talent for innovation. Söderström called it the beginning of AI's role in development, not the end.
Conclusion: The Future of Coding Is Here
Spotify proves AI can lead software engineering at scale. Top engineers thriving without writing code redefines productivity. As competitors watch, expect this shift to ripple across tech, prioritizing judgment over manual labor.
This evolution promises faster innovation for users worldwide. Spotify's story shows AI amplifies human potential, paving the way for the next era of development.
This represents a truly exciting milestone in software engineering! 🚀 Spotify's approach demonstrates how AI can amplify human potential by freeing top engineers from repetitive coding to focus on strategic architecture and innovation, delivering impressive results with 50+ features shipped, €4.5B revenue, and 38M new users. However, there's a critical caveat that affects us as end users: if the human review process isn't rigorous enough, we could face serious production issues like subtle bugs, security vulnerabilities, performance degradation, service outages, data corruption, or even privacy breaches. 🔍 The article's mention of engineers reviewing AI-generated code from their phones during commutes is convenient but concerning, sophisticated bugs require deep focus to catch, and rushing reviews could mean users experience broken features, payment errors, or accessibility problems that slip through automated testing. ⚠️ The innovation is genuinely groundbreaking, but the speed gains only benefit users if review quality remains proportional to production risk, making rigorous human oversight the non-negotiable foundation of this AI-first approach