Scaling code health with always on rewrites and AI at Spotify
summary
Spotify describes a system that applies Java conventions across 2,900 services and 190,000 repos using OpenRewrite and Error Prone, rolls changes via a Kubernetes-based Fleet Shift that has produced 2.5 million PRs merged, and uses an LLM-backed background agent called Honk that runs full CI (Maven, Bazel, test suites) before opening PRs. Cohort rollouts, Auto Merger and Fire Watch enable safe auto-merges, and Log4j was remediated across ~80% of services within a week.
AI-written summary · about
chapters (9)
- 0:00Introduction and speakers
- 1:29Spotify's engineering scale
- 3:57Java conventions at scale
- 9:09Golden path and tech standards
- 12:01Fleet Shift platform overview
- 17:18Auto merging and trust at scale
- 23:58Honk AI coding agent
- 30:50Honk verification and future
- 37:11Honk in Slack and demos
related talks
49mMicroservices at Spotify
27mBalancing autonomy, technical alignment, & speed through strategic collaboration
31mThe Rise of Internal Developer Portals and the Future with AI
36mAdopting gRPC at Spotify
48mWeeding Your Micro Service Landscape
01h 16mAutomating Large Scale Refactors with Parallel Agents
28mNavigating the Transition: SRE Challenges and Highlights
46mModelling Microservices at Spotify