Author here. I built this because after years of watching teams spend more time fighting Kubernetes YAML, service meshes, and Spring Boot startup than writing business logic, I wanted to see if the JVM itself could be the platform.

Aether is a standalone distributed runtime for Java 25. You write "slices" (small focused handlers), package them as JARs, and Aether handles clustering, deployment, routing, rolling updates, auto-healing, and auto-scaling. No containers, no sidecars, no orchestrator.

Under the hood: Rabia consensus for replicated state, consistent hash ring DHT for artifact storage and distributed cache, blueprint-based deployment with traffic shifting, and CPU-based auto-scaling with optional ONNX-based predictive scaling.

Performance: 4K RPS at ~0.5ms latency (localhost, Mac M1 Air), 10K req/s distributed. Survives rolling restarts with ~100% request success.

Just released 0.17.0 with pub-sub messaging, scheduled invocations (cron + interval), DHT anti-entropy repair, and cluster event streaming.

59 features complete, 80+ E2E tests, 427+ unit tests. Not claiming production-ready for unsupervised use yet -- no RBAC (coming in 0.18.0), no KV-Store backup. Ready for controlled pilots.

Core library and tooling are Apache 2.0. Aether runtime is BSL 1.1 (free for internal use at any scale, converts to Apache 2.0 in 2030).

1-min demo: https://www.awesomescreenshot.com/video/49013877?key=a0b1105...

Happy to answer questions about the architecture or the "JVM as platform" approach.