Coding agents are changing how software gets built, tested and operated.

Gradle Is Going Agentic

🎯 Featured Article

🤖 Gradle Is Going Agentic Gradle is adapting its build tooling for coding agents.

  • Agent-driven builds are becoming part of the normal development loop.
  • Agentic Gradle combines agent skills with new build-tool features.

⏩ TL;DR (Quick Recap)

  • Build tools are adapting to agents as they enter everyday development workflows.
  • Java and Spring continue improving startup, security, configuration and AI workloads.
  • Agentic development is reshaping testing, code review, migrations and tool selection.
  • Production engineering still depends on careful validation, observability and controls.

☕ JVM Corner

📦 Module Imports Java 25 introduces module imports, allowing types from an entire module to be imported with a single declaration.

  • Module imports extend the convenience of wildcard imports to modules.
  • The feature reduces import declarations when module-wide access is appropriate.

🐢 Why is Arrays.fill 265 times slower on G1GC? A JVM profiling deep dive investigates why Arrays.fill can behave dramatically differently under G1 garbage collection.

  • JVM barriers can materially affect seemingly simple array operations.
  • Diagnostic JVM flags help identify the mechanisms behind performance differences.

📱 Native is now the future of mobile at Shopify Shopify is moving mobile development toward Swift and Kotlin as coding agents reduce the cost of maintaining native applications.

  • Coding agents changed the economics of mobile development.
  • Shopify sees native platforms as a better fit for its current needs.

🪟 The Evolution of WSL Support in JetBrains IDEs — JetBrains is consolidating several generations of WSL integration into a more consistent IDE experience.

🍃 Spring Updates

🔐 Post-Quantum Cryptography in Spring Boot: Four Patterns You Can Ship This Sprint Spring Boot applications can begin adopting post-quantum cryptography through JDK 24’s standard ML-KEM and ML-DSA APIs.

  • JDK 24 exposes ML-KEM and ML-DSA through the standard JCE API.
  • PQC migration can begin before managed cloud TLS support becomes universal.

🧩 Vertical slices in Spring Boot, and how to test them Vertical slices organize Spring Boot code around features rather than technical layers, changing both implementation and testing.

  • Features become the primary unit of application organization.
  • Tests can follow the same feature boundaries as production code.

🧠 Spring AI Recipe: Running Against an in-JVM Model Spring AI can run local model inference directly inside the JVM, removing the need for a separate model server.

  • ModelJars allow models to run alongside the application.
  • Inference can remain inside the same JVM as Spring AI.

🚀 Improving First Request Latency in Java Spring Application Spring AOT features can improve more than startup time by reducing work that otherwise appears on an application’s first request.

  • First-request latency can differ substantially from subsequent requests.
  • Recent AOT improvements can move work earlier in the application lifecycle.

🗜️ Spring AI Recipe: Summarizing Chat Memory Spring AI can summarize conversational memory to preserve important context without carrying the entire conversation forward.
⚙️ Managing External Configurations with Spring Cloud Config Spring Cloud Config provides a central mechanism for managing external configuration across distributed environments.

🔍 Extra Reads

Agentic Engineering
🧐 Maybe We Shouldn’t Be Reviewing All This Code — AI may expose that traditional code review is solving problems that could be addressed elsewhere.
🧰 Which tools do Claude Code, Codex and Cursor choose? — A study of 16,893 sessions examines how three coding agents select tools in real repositories.
🧪 Hangar Harness / Model Tests — Model and harness combinations are tested to understand which setups produce better agent results.
🔄 We Let AI Agents Rewrite a 92M-Message-a-Day Service in Go. Zero Incidents. — An agentic Node.js-to-Go rewrite achieved lower resource usage after being constrained by a purpose-built test harness.
💸 RTK reports huge token savings, but our cost benchmarks disagree — Terminal-output compression can reduce agent input tokens without necessarily reducing total AI costs.
🧬 Open-weight models pulled ahead — where the work is hardest — A clinical-regulatory benchmark found an open-weight model leading on the difficult TLF programming task.

Web
🎨 Tailwind Labs is joining Shopify — Tailwind Labs is joining Shopify, which will provide a long-term home for the widely adopted framework.
🌐 Chrome Is Switching Off unload: What Breaks and How to Fix It — Chrome’s unload shutdown affects applications that rely on the event for state, locks or server notifications.

Originally posted on marconak-matej.medium.com.