Netflix’s Journey to Confident Automated Changes
summary
Netflix runs automated dependency updates across ~5.6k repositories and ~3.6k Java applications (40k active libraries, ~1,500 Java engineers) and moves from manual PRs to a zero-touch workflow that uses OpenRewrite for deterministic migrations and LLMs for messy edge cases. The system adds symbol-level indexing and blast radius scoring to rank risk, surface downstream callers, and enable auto-merging low‑risk upgrades; a Confidence Chimp end‑to‑end pipeline runner validates build→deploy before merging and teams see health via Gold to Stone Age scorecards and a testing‑pyramid driven CI strategy.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:50Challenge of frequent updates
- 10:02GenAI tools and MCP platform
- 22:26Building confidence in changes
- 33:20Testing insights and validation
- 46:21Current investments and blast radius
- 49:13Key learnings from journey
- 52:44Takeaways and conclusion
related talks
25mJourney to Automated Releases
42mDependency Updates: Before and After
43mNetflix + Gradle A Journey in Developer Productivity
20mDeveloper Productivity in the Age of GenAI
39mMoving fast and staying aligned how automation powers progress
17mRelationship Therapy: Keeping Developers and Security Teams Happy
29mHow we turn data teams into better engineers at Netflix
28mTowards Automated Updates of Software Dependencies