Debugging production issues: When something breaks at 3 AM in production and you need to figure out why, AI can suggest things to check, but you're the one reading logs and understanding system behavior.

Understanding business requirements: Translating "we need a feature that helps users discover music based on their mood" into technical specifications requires human judgment.

Code review: Evaluating whether a solution is maintainable, performant, and aligned with system architecture - AI can flag obvious issues, but judgment calls still require engineers.

So what's really happening at Spotify? Probably something like this: engineers describe what they want to build using natural language or high-level specifications. AI generates implementation code. Engineers review the output, test it, refine it, and ship it.

The productivity gain is real - writing less boilerplate means more time for the hard problems. But calling this "engineers not writing code" is like saying authors don't write anymore because they use word processors instead of typewriters.

The tool changed. The core skill didn't.

I built a fintech startup. I wrote a lot of code. I also copy-pasted a lot of Stack Overflow answers and modified template projects. If AI had existed then to automate that part? Absolutely, I would have used it. It would have made me more productive. But it wouldn't have changed the fundamentally creative work of figuring out what to build and how it should work.

The danger of claims like Spotify's is that they fuel the "AI will replace programmers" panic without acknowledging what's actually happening. AI is a productivity tool for engineers, not a replacement for engineering.

That said, the implications are still significant. If AI makes engineers 2-3x more productive, companies need fewer engineers for the same amount of output. That's a real labor market effect. Junior engineers who spend most of their time on boilerplate might find fewer opportunities.

But the idea that Spotify's best engineers - the ones making architectural decisions, debugging complex distributed systems issues, and designing scalable data pipelines - have been replaced by AI? That's not what "haven't written code since December" means.

It means they've changed their workflow. They're still engineering. They're just using better tools.

The technology is impressive. The question is whether the framing is honest. Because "AI helps engineers be more productive" is a very different claim from "AI has replaced coding," and Spotify seems to be marketing the latter while delivering the former.

4 days ago