Agentic development is moving from code generation toward systems, platforms and engineering judgment.
🎯 Featured Article
🛩️ JetBrains Air: Building a System of Products for Agentic Software Development Multi-surface system connecting products for developers, teams and organizations around agentic software development.
- Connects multiple products across development workflows
- Extends agentic development beyond individual coding tools
⏩ TL;DR (Quick Recap)
- Agentic development is expanding from coding assistants into broader engineering workflows
- Java continues evolving through JVM performance, Quarkus, Kotlin and developer tooling
- Spring AI enables increasingly lightweight workflows for chat and MCP applications
- Engineering judgment remains critical as AI-generated code becomes more capable
☕ JVM Corner
🔗 MCP goes stateless, and Quarkus already implements it MCP’s move toward stateless sessions simplifies distributed deployments by removing sticky-session requirements.
- Eliminates session affinity for distributed MCP servers
- Simplifies horizontal scaling with standard HTTP requests
⚙️ Just-In-Time Compilation for Java Performance: Recent and Ongoing Improvements JIT compilation remains central to Java performance while adapting to language, deployment, hardware and JVM changes.
- Supports projects including Valhalla, Leyden and Panama
- Adapts compilation strategies to evolving hardware
🧩 Decapsulation: Breaking Java Strong Encapsulation An exploration of techniques for reaching JDK internals through reflection, agents, FFM, bytecode manipulation and more.
- Demonstrates multiple approaches to JDK internals
- Covers reflection, agents, FFM, Unsafe and bytecode techniques
📝 My love-hate relationship with Java formatters A long-time Java developer examines formatting conventions and introduces a formatter project based on years of experience.
📊 The State of Kotlin in 2026 Report JetBrains reviews Kotlin’s first fifteen years, including ecosystem maturity, adoption, multiplatform development, and AI.
🌐 The quarkus.io site is now built with Quarkus Roq Quarkus has migrated its website from Jekyll to Roq, a Java-based static site generator built with Quarkus.
🍃 Spring Updates
🔄 Spring Boot 3 to 4 Migration: Fixing Every Break A migration walks through upgrading a Spring Boot 3.5 application to Spring Boot 4.x.
- Covers compiler, test and runtime migration failures
- Explains why individual compatibility problems occur
💬 Spring AI Recipe: A No-Build Chat Loop with JBang Spring AI can run a chat loop directly from Java source with JBang, without creating a Maven or Gradle project.
- Uses JBang directives instead of a build project
- Runs Spring AI directly from Java source
🔌 Spring AI Recipe: No-Build MCP with JBang Spring AI MCP servers and clients can run directly from Java source with JBang, removing the traditional build step.
- Builds MCP components directly from Java source
- Supports STDIO MCP workflows without Maven or Gradle
⚖️ Spring AI Recipe: Making AI Judgments with Jev Jev and Spring AI use AI models for focused judgments that return structured results applications can consume directly.
- Produces structured results for application logic
- Targets focused classification and judgment tasks
🔍 Extra Reads
Agentic Engineering
🤔 I don’t like LLMs — Examines the productivity potential and risks of LLM technology.
🏚️ Brownfield Agentic Engineering — Agentic development in legacy systems requires exposing hidden constraints before making changes.
🧠 Coding is NOT solved — A challenge to the claim that AI has solved software development and reduced engineering to taste.
🔭 What I believe about the future of software development — Building software will matter more as writing code becomes increasingly automated.
AI Models and Tools
⚡ Claude Sonnet 5.5 — Faster, lower-cost model aimed at everyday development and professional tasks.
🧮 GPT‑6.1 Sol — Lower-cost model targeting agentic coding, computer use and professional workflows.
💠 Gemini 4 Argon: our next era of frontier intelligence — Frontier model targeting complex software, enterprise and cybersecurity workflows.
🟣 Introducing dots — Always-on agents designed to handle ongoing work on behalf of users.
🛠️ Pi 1.0 — Hardened, minimal, extensible agent harness developed through broad user feedback.
Engineering Practice
🗂️ From documenting decisions to questioning them — An AI agent moves beyond compiling decision records toward challenging the decisions they document.
🔐 I don’t like passkeys — A critical perspective on passkeys, their adoption and the user experience.
🧱 Back to Coupling and Cohesion — How software design principles such as coupling and cohesion can be applied through practical metrics.
🔗 Don’t couple your Go code to GitHub — Why Go packages should avoid coupling their import paths directly to Git hosting providers.
📚 Thirteen years of blogging — Reflects on thirteen years of technical blogging and how developer discovery has changed.
💻 Ten Lines Of Code That Changed My World — A personal collection of ten memorable pieces of code and the reasons they mattered.
Originally posted on marconak-matej.medium.com.