From AI-assisted debugging to agentic workflows and JVM performance breakthroughs
🎯 Featured Article
🧠 Teaching an AI Agent to Debug Flaky Tests
A dive into how AI Skills, deterministic workflows and developer tooling can help agents tackle one of the most frustrating problems in software engineering: flaky tests.
⬢ Explains how reusable AI Skills act as domain-specific operational playbooks for agents.
⬢ Shows how deterministic debugging workflows outperform generic prompting for engineering tasks.
⏩ TL;DR (Quick Recap)
⬢ AI engineering is shifting toward structured orchestration, governance and workflow-driven systems.
⬢ Java and JVM ecosystems continue optimizing runtime efficiency.
⬢ Spring and JVM tooling ecosystems are focused on enforcing architectural correctness at build time.
⬢ Developers are reassessing the trade-offs of agentic coding and long-term cognitive debt introduced by AI.
☕ JVM Corner
☕ MCP in the Java World: Bringing Architectural Strategy to LLM Integrations
Explores how Model Context Protocol introduces governance and architectural discipline to LLM integrations.
⬢ Positions MCP as an architectural abstraction instead of simple tool-calling glue.
⬢ Highlights loose coupling, versioning and governance as critical AI concerns.
⚡ Enabling reflection-free Jackson serializers by default
Quarkus introduces build-time generated Jackson serializers to eliminate reflection overhead and improve runtime efficiency.
⬢ Reduces serialization costs by replacing reflective access.
🐳 The Road to Docker Official Images for Java: The Azul Zulu Story
A behind-the-scenes look at how Azul Zulu became an official Docker image distribution.
⬢ Explains the operational and security requirements behind Docker Official Images.
⬢ Highlights the growing importance of trusted JVM container distributions.
🤖 JobRunr Introduces ClawRunr, an Open-Source Java AI Agent
ClawRunr brings persistent scheduling, retries and orchestration capabilities into Java-based AI agent execution.
⬢ Focuses on long-running reliability concerns often ignored in AI agent design.
🚀 Faster Startup on IBM Semeru with OpenJ9 Shared Classes Cache
Quarkus explores faster startup techniques for OpenJ9-based JVM distributions using Shared Classes Cache.
⬢ Improves startup performance for OpenJ9 deployments.
🍃 Spring Updates
🕸️ Spring AI Recipe: Building a Graph-Based Agentic Workflow
Demonstrates how graph-based orchestration can make AI agents more predictable and maintainable.
⬢ Treats workflows as structured decision graphs instead of fully autonomous loops.
⬢ Balances flexibility with operational control in agentic systems.
🛡️ Spring Boot Best Practices That Should Fail Your Build
Argues that architectural conventions should be enforced automatically rather than documented informally.
⬢ Promotes build-time validation of common Spring anti-patterns.
⚙️ Spring Boot, Micronaut and Quarkus with Mill
Explores how the Mill build tool improves developer experience for modern JVM frameworks.
⬢ Highlights faster incremental builds and simplified workflows.
🔍 Extra Reads
AI Engineering & Developer Workflow
🧩 The Code Was Always the Door — Explores how senior engineers increasingly guide AI systems rather than manually producing every implementation detail.
🎭 Appearing Productive in The Workplace — Examines how AI tooling can amplify performative productivity rather than meaningful output.
⚠️ Vibe coding and agentic engineering are getting closer than I’d like — Simon Willison reflects on the convergence between exploratory AI-assisted coding and fully agentic engineering workflows.
🧠 Agentic Coding is a Trap — A cautionary perspective on cognitive atrophy and over-reliance on AI-generated implementation work.
Tooling & Performance
⚡ Speeding up interactive rebase in JetBrains IDEs — Details low-level Git optimizations that significantly improve interactive rebase performance in large repositories.
🚗 OurCar: What I Learned Making an App for my Family — A practical reflection on building a real-world Flutter app.
Originally posted on marconak-matej.medium.com.