Stripe builds continuous Java upgrades with Minions fixing CI failures, reaching 93%+ of services on JDK 17 within a week of rollout
Stripe's Java team built AutoJDK, which reads the live Bazel build graph to work out which targets can move to a newer JDK, generates bounded reviewable PRs, and tracks blockers. Stripe's Minions coding agents handle the fixable failures, using fresh git checkouts in sandboxes, and can be launched from Slack. For large upgrades an agent skill runs upgrades across all partitions, monitors and fixes CI, and cherry-picks commits into PRs covering around 10,000 targets. LogScale alerts catch runtime failures for targeted rollback.
The numbers (as reported)
| JDK 17 adoption | 93%+ of services within under a week of rollout |
|---|---|
| PRs merged | 160+ in three months |
| Legacy JDK footprint | Down by roughly 25K targets; all Java 8 and most Java 11 services gone |
| Largest PRs | About 10K targets upgraded per PR |
Facts
- Company / subject
- Stripe
- Tools
- Unstated
- In-house tools
- AutoJDK, Minions
- Codebase
- brownfield
- Work type
- Upgrade / modernisation
- Stack
- Java, Scala, Bazel
- Published
- 2026-05-27
- Last checked
- 2026-11-02
What the source claims Practitioner (own blog/paper)
Claimed by the builder, writing about their own work. Not checked independently.
- JDK 17 adoption reached 93%+ of services in under a week of rollout 93%+
“In under 1 week of rollout progress, service-level adoption reached 93%+ on JDK 17”
- 160+ PRs merged and the legacy-JDK footprint cut by roughly 25K targets, with Java 8 services eliminated 160+ PRs; ~25K targets
“Shrunk the legacy-JDK footprint, specifically JDK 8 and 11, by roughly 25K targets”
Caveats
Much of the work is deterministic build-graph automation; agents handle a subset of failures. The post doesn't say how many fixes came from Minions or which model they use.
More stories by tool: Unstated
Related topics: Upgrades and modernisationBrownfield codebases