Spring AI 2.1.0-M1 models an assistant turn as an ordered list of parts. Here is what the model looks like, how streaming marks partial parts, and where reasoning is silently lost today.
A Spring Boot 4.1 service started 3.9 s on a plain JDK 25 and 2.0 s with the JDK AOT cache, from one extra flag. Measured against AppCDS, Spring AOT and a GraalVM native image, with 24 ways the cache gets silently rejected (rebuilt jar, copied files, ZGC, compact headers), the Docker consequences, and what is and is not known about Netflix's production use.
Measured jlink results on Temurin 25 for a real Spring Boot 4.1 app: which modules it needs (jdeps), what each jlink option saves, why java.desktop costs 15 MB, a failure when you remove it, and a JMOD-based custom runtime of 42 MB for a minimal modular program. Docker was not available, so no image sizes are claimed.
What FIPS 140-3 means for a Java service, what Oracle Jipher 20 is and requires, and a runnable Spring Boot 4.1 service with HTTPS and an ES256 JWT that shows which JCA provider serves every algorithm and which TLS settings Oracle's guide asks for. I could not run Jipher itself, and the post says so.
Oracle plans to move Oracle JDK 21 updates from the No-Fee license to the OTN license starting with the 20 October 2026 patch update. What that does and does not mean, how to check which JDK you actually run, the support dates of the main OpenJDK builds of 21 and 25, and a migration checklist.
Runnable post-quantum cryptography in Java: ML-KEM and ML-DSA round trips with real key, ciphertext and signature sizes next to RSA and ECDSA, the hybrid X25519MLKEM768 TLS 1.3 handshake that JDK 27 enables by default, the failure you hit if you enable only a non-default hybrid group, and measured speed on JDK 25 versus 27. Three downloadable Java files.
The same Spring Boot 4.1.1 service on JDK 25 and JDK 27, with compact object headers opted in and out. The footprint win is real and repeatable (about 24 MiB less live heap); throughput, p99 and startup are not distinguishable on a small box. Includes the numbers, the noise, and a flag that gets the memory win on JDK 25 today.
JEP 545 and JEP 546 aim to make ZGC pick its own heap size and start faster in JDK 28. Both are only proposed for JDK 28 and the flag is not in any build yet, so this post explains the design, measures today's startup and warmup baseline on JDK 25, 27 and 28 early access with a Spring Boot 4.1.1 server, and gives you a harness to rerun when the flag lands.
JDK 28 adds a small JSON parser and generator to the JDK itself (JEP 540, jdk.incubator.json). Compiled and run on JDK 28-ea+19: the tree model, typed access, error messages, pattern matching, number and duplicate-key rules, and when to still use Jackson.
Run a DistilBERT sentiment model from Java with ONNX Runtime and DJL: tokenizing, identical answers by two routes, batching, a Spring Boot endpoint, and measured latency.