In a recent LinkedIn post, Linas Beliūnas discusses a significant shift in software development, highlighting how Spotify is leveraging artificial intelligence to automate core coding tasks. Beliūnas points to Spotify’s revelation that top engineers have not written code since December, instead delegating tasks to an internal AI system named “Honk,” powered by Claude.
According to Beliūnas, Spotify’s co-CEO Gustav Söderström explained on the Q4 2025 earnings call how engineers now use AI for coding, even remotely. The process involves Claude writing code, deploying new builds, and sending them for review, often before engineers even arrive at the office.
“Claude writes the code. Deploys a new build. Sends it back for review. All before they even reach the office.”
Beliūnas notes that this AI-driven approach has coincided with Spotify shipping over 50 product updates in 2025, including features like AI-powered Prompted Playlists and “About This Song.” However, he emphasizes that the true significance lies not just in AI writing code, but in the evolution of the engineer’s role.
The Bottleneck Shifts to Judgment
Linas Beliūnas argues that the primary bottleneck in software development is moving from the act of typing code to the act of making critical judgments. He posits that engineers are not becoming obsolete but are transforming into orchestrators of AI systems.
“Which yet again proves that engineers aren’t disappearing. They’re becoming orchestrators 🤖”
This transformation, Beliūnas suggests, is crucial for companies aiming to thrive in the AI era. The ability to structure and manage AI systems that can write code safely and effectively is becoming a more valuable skill than traditional coding prowess.
Building Proprietary Datasets: The Uncommoditizable Asset
Beyond the automation of code, Beliūnas highlights another critical aspect of Spotify’s strategy: the development of a proprietary taste dataset. He explains that understanding user preferences, such as music tastes, is not a matter of factual recall but requires analyzing behavioral data at scale.
Linas Beliūnas elaborates on this point by stating:
“Taste isn’t Wikipedia. It’s behavioral data at scale.”
He contrasts this with easily commoditized aspects, suggesting that companies that can harness and interpret this unique behavioral data will possess a significant competitive advantage. This dataset allows for more personalized user experiences, catering to diverse preferences across different regions and demographics.
Navigating the Challenges of AI Implementation
While acknowledging the potential of AI in accelerating product development, Beliūnas also addresses the concerns raised by skeptics. He notes the potential for accumulating technical debt if speed is prioritized over quality, citing issues like app bugs and shuffle complaints.
However, Beliūnas remains optimistic about the long-term implications. He concludes that the future leaders in the AI age will be those who develop superior AI systems, rather than simply mastering prompt engineering.
“The companies that win in the AI age won’t have better prompts – they’ll have better AI systems.”
This perspective underscores the ongoing evolution of technology and the workforce, emphasizing strategic system building and data mastery as key differentiators for future success.
📝 About This Content
This article is based on insights shared by Linas Beliūnas on LinkedIn.
📅 Originally posted on February 15, 2026 | View original post on LinkedIn →