Spotify: AI doubled development speed, quality checks became the...

Spotify: AI doubled development speed, quality checks became the bottleneck

Spotify's engineering blog reported that AI increased the number of merged changes from about 8,100 to 17,000 per year, while major incidents were not caused by AI code but by reviews, tests, and rollouts failing to keep up with the volume growth.

Compiled by Sergey KostenchukPublished 2026-09-20Updated 2026-09-20

2026-09-20 Coding

💻 Spotify: AI doubled development speed, quality checks became the bottleneck

In Spotify's engineering blog (author — Tyson Singer, SVP): merged changes — 17,000 per year instead of 8,100, quality work — 31% instead of 27%. Major incidents in the month did not cause AI code, but reviews and tests did not keep up with the growth.

🌍 The thesis of "AI slop" was not confirmed by incidents: the industry is strengthening reviews, tests, rollbacks, and observability, rather than banning AI code. AI demand for CPU/GPU consumed free capacity — after the May failure, edge capacity reserve was doubled.

👤 Report of a service with about 777 million users and about 3,000 services: speed doubled, but quality control had to be rebuilt. Check your pipeline: change volume control, batch priorities, rollback reserve.

Source 1: https://engineering.atspotify.com/2026/9/ai-changed-how-spotify-builds-what-we-learned-and-fixed-about-quality-at-higher-velocity Source 2: https://www.faros.ai/research/ai-acceleration-whiplash

Sources