Contents
JCON
▪️A Big Screen Experience for Java Developers & Architects
CORE JAVA
▪️A Big Screen Experience for Java Developers & Architects
▪️Effective Pattern Matching 2026 Edition
▪️Data Oriented Programming with Java
▪️Java 26 adopts HTTP/3 with the Evolution of the HttpClient
▪️Java 26 GC: Making Data Streaming Faster and More Reliable in Containers
▪️1BRC – The First 80%
▪️Building GPU Accelerated LLM Inference Libraries with TornadoVM
▪️Reactor Has Died, Long Live REACTOR!
API & FRAMEWORKS
▪️The FFM API: OpenJDK Changed the Game for Native Interactions and #JavaOnRaspberryPi
▪️Talk to Your Data: Natural Language Data Access in Java
GENAI
▪️Build Vector Database Apps with Pure Java
▪️High-Performance Vector-Search Grids with Java
▪️LLMs for JVM Code in 2026: Which Models to Trust, Which Benchmarks to Believe
▪️Java 26 + Agentic Microservices: Cloud Performance Re-Imagined for FinTech
▪️Building Production-Ready AI Agents with Java and Spring AI
▪️A2A Building Interoperable AI Agents with Java
▪️The 5 Knights of the MCP Apocalypse
▪️Petabyte-Scale AI Memory with Serverless Java
ARCHITECTURE & MICROSERVICES
▪️Zero Migration Java – Staying Current Without Breaking Your App
▪️Enterprise Release Evolution
▪️Capability Mesh: Re-Architecting for the Agentic AI Future
CI/CDI
▪️Bye-Bye Jenkins: Why We Migrated our Pipelines to Gitlab CI
PROJECT MANAGEMENT
▪️AI’s Impact on Team Dynamics
SECURITY
▪️PQC – It is Time to Act
▪️The Myth of Stability: Java’s Software Supply Chain After Log4Shell
JVM LANGUAGE
▪️Kotlin kontra Java – Part 1 – Ecosystem
Details
Page count: 140
Authors: Cay Horstmann, Birgit Kratz, Wanderson Xesquevixos, Varsha Sharma, Shubham Goel, Rene Schwietzke, Michalis Papadimitriou, Marcin Chrost, Frank Delporte, Marco Belladelli, Richard Fichtner, Gerald Kammerer, Artur Skowronski, Sibasis Padhi, Yuriy Bezsonov, Sascha Möllering, Loïc Magnette, Jonathan Vila, Markus Kett, Grace Robinson, Laura Cowen, Matthias Zieger, Makan Sepehrifar, Lukas Pradel, Erik Pronk, Sebastian Hempel, Steve Poole, Richard Gross
Editorial
Java and the AI Renaissance:
Late to the Race, Built to Lead
Java and the AI Renaissance: Late to the Race, Built to Lead
The AI landscape is shifting at a breathtaking pace, and one recent announcement has made that unmistakably clear. With TurboQuant, Google introduced a breakthrough in model efficiency by compressing neural representations into ultra-compact formats, drastically reducing memory footprint and compute requirements. The immediate impact was striking – not only in performance benchmarks, but even in financial markets, where shares of major AI hardware vendors briefly reacted to the implication that less hardware might be needed for the same workloads. TurboQuant signals a turning point: optimization is no longer incremental, it is architectural.
Yet while TurboQuant addresses short-term memory and inference efficiency, the larger challenge remains unresolved. Long-term AI memory – the persistent, growing knowledge layer behind intelligent systems – still operates on architectures that were never designed for it. This is where a new generation of Java-based technologies is beginning to redefine the landscape.
With EclipseStore 4, the Java ecosystem now has a fully native, pure Java vector database engine. For the first time, Java developers can build GenAI applications entirely within the Java runtime, without relying on external vector databases or polyglot architectures. Combined with Eclipse Data Grid, which extends this model into a distributed in-memory vector search grid, developers can scale these applications seamlessly across clusters. Both projects are fully open source, developed under the Eclipse Foundation by a team that has spent over a decade building core technologies such as Eclipse Serializer and EclipseStore. Every feature is transparent, community-driven, and built with a clear commitment to the Java ecosystem.
At the same time, Cyrock.AI introduces a complementary perspective on the problem. If TurboQuant optimizes the efficiency of models, Cyrock.AI focuses on the efficiency of knowledge itself. Its Knowledge Fabric architecture reimagines how long-term AI memory can be structured, accessed, and scaled. By applying principles similar to serverless computing – allocating resources only where data is actively used – it enables a fundamentally more efficient way to handle petabyte-scale knowledge systems. The implications are significant: not just lower costs, but the possibility of truly persistent, continuously evolving AI memory.
Taken together, these developments point to a broader transformation. Software is moving from being purely code-driven to being shaped by context. Java systems have long been built on deterministic logic, where behavior is explicitly defined and controlled. That foundation remains critical, but it is no longer sufficient on its own. Modern systems increasingly operate by interpreting situations, combining data, and dynamically determining outcomes.
This shift introduces a new paradigm: from direct control to guided autonomy. Developers are no longer only writing execution paths; they are defining boundaries, constraints, and environments in which systems can act. The focus moves from specifying every step to ensuring that decisions are made within acceptable limits.
At the center of this transformation lies context. AI systems depend on access to relevant, structured, and trustworthy information. Data is no longer passive storage – it becomes an active component of system behavior. Technologies such as vector search, graph traversal, and distributed in-memory data grids are not just performance optimizations; they are the foundation of how modern systems reason.
This also raises new questions of governance and responsibility. As systems become more autonomous, ensuring transparency, traceability, and control becomes essential. The challenge is not only to build powerful systems, but to build systems that can be trusted.
Through all of this change, one constant remains: Java itself. Its stability, maturity, and performance make it uniquely suited for this new era. Rather than being replaced, the Java ecosystem is expanding – embracing AI, distributed systems, and new data paradigms while preserving its core strengths.
What we are witnessing is not just the evolution of tools, but the evolution of software itself. From code to context, from deterministic execution to intelligent systems, the role of developers is changing. The question is no longer only what systems should do, but how they should think – and what they should be allowed to decide. This issue explores that transformation.
Before closing, a brief personal note. As JAVAPRO we’re neutral related to technologies, vendors and products. Exceptionally, I like to highlight projects that my team and I have been building for more than a decade: Eclipse Serializer, EclipseStore and Eclipse Data Grid. These technologies are the result of continuous iteration, shaped and refined through close feedback from the Java community. Today, all of those ideas and concepts come together into a powerful, fully open-source pure Java GenAI stack built specifically for Java developers – created with passion and a clear goal: to contribute something meaningful that helps the Java community and strengthen Java’s role in the AI era. We are especially proud to introduce these frameworks at JCON EUROPE 2026, and I hope they provide real value to you in your GenAI projects.
A big thank you goes to our fantastic authors of this issue as well as to all the authors who write enthusiastically for JAVAPRO online – without you JAVAPRO would not be possible.
If you are part of a Java User Group, take advantage of the free JUG tickets, connect with your local Java User Group, and get involved. Let’s strengthen the Java ecosystem together and meet in person at JCON. I wish all of you an inspiring and unforgettable event.
