Boris sat down with Spotify VP of Engineering Niklas Gustavsson.
Spotify ships 4,500 production deploys a day, and 73% of PRs are now AI-assisted.
The company suffered three service disruptions over four weeks.
Boris sat down with Spotify VP of Engineering Niklas Gustavsson.
Spotify ships 4,500 production deploys a day, and 73% of PRs are now AI-assisted.
Users in the replies largely dismissed Spotify's claims of 4,500 daily deploys with 73% AI-assisted PRs as vanity metrics or slop that ignore the app's poor quality and reliability problems.
No Digg Deeper questions have been answered for this story yet.

Watch the full interview here:
https://www.youtube.com/watch?v=9DHZLw5653E

Niklas keeps 5 to 10 Claude sessions running in tmux, one per git worktree, agents working in the background.
All of it inside a 20M+ line monorepo. He expected agents to struggle at that size, but it's worked well.
Coincidentally, 3 out of the last 4 weeks I could not include a link to Spotify when releasing my podcast episode, b/f Spotify was down.
2x an outage in publishing the podcasts, 1x the complete web player down
Have not experienced such poor reliability from Spotify in forever
Boris sat down with Spotify VP of Engineering Niklas Gustavsson.
Spotify ships 4,500 production deploys a day, and 73% of PRs are now AI-assisted.

Spotify's migration codemods grew into thousands of lines of edge cases. Code has too much API surface for static rewrites. Early LLMs barely did better.
Adding a judge took PR success from ~25% to 80%.

All of this leans on verification, the single most important thing when agents are used and the place most companies underinvest
Spotify rebuilt their test automation around it so engineers can confidently guide and supervise agents, rather than manually execute repetitive tasks

@ClaudeDevs Product looks the same for the last 10 years. What are they deploying?

@ClaudeDevs 4500 a day? How much of that reachs the end user? 😅

@ClaudeDevs 4500 deploys a day to play music?
to play music

@ClaudeDevs What are they pushing with 4,500 deploys ?A DAY? A DAY? Can someone enlighten me?

@ClaudeDevs 4500 deployment, and most importantly nobody likes them. It's not how many you ship, its always what you ship ?

@ClaudeDevs The 73% PR number is flashy, but the useful detail is lower down: a judge took migration PR success from ~25% to 80%. That’s the pattern I’d copy first. Not more agents. Better verification around one boring migration.

@ClaudeDevs 4500 deploys per day? WTH are they shipping?

@ClaudeDevs

@ClaudeDevs Are these deploys in the room with us right now?

@ClaudeDevs No major announcement for Plan users in a bit… You’ve either gotten us too used to it or you are preparing something great. Hope to see Sonnet 5 -> Opus capabilities at a lower price, even lower than now, Opus 5 Fable capabilities in coding and long-running tasks, Fable 5.1…

@ClaudeDevs You meant 4.5 milion PRs a day for they music app that already exist for 10 years

@ClaudeDevs 4500 releases a day? From a user perspective, if you’d ask me about what has changed in the app since last year… I don’t know lol. It feels stale for years. Podcasts? Dailymix? Something feels wrong

@ClaudeDevs @trq212 How come only 73% is AI-assisted? Besides the songs, podcasts, and their S3 paths, what doesn’t have any AI in it? This is nonsense.
To be clear, I am not implying that Spotify's reliability problems are to do with eg Claude Code. But reliability seems to be far more problematic than before from my POV
Eg outage a month ago:
While I appreciate that the eng problem of publishing video (in this case: video podcasts) is not an easy one... it has been solved very well, by YouTube, X, many others...
Meanwhile an hour later, Spotify cannot publish my podcast episode, and their target is "a few hours"....

@ClaudeDevs Is this amount of production deploys a good thing?