When AI writes part of the code, where exactly does judgment still stay human? I wrote about the responsibility gap, oversight design, provenance, and why vague human oversight is not enough. buff.ly/EIKXfsw

I wanted a sample that kept the ownership boundary visible instead of quietly inlining the KB into helpdesk code. This one uses Quarkus, LangChain4j, and A2A to split triage from remediation in a way teams can actually run. buff.ly/pciu2Lx #Java #Quarkus

AI changes authorship faster than it changes accountability. I wrote about code liability, the verification gap, provenance, and why enterprise Java teams should stop treating AI-written code like free speed. buff.ly/RX1erUQ #AICoding #Java

I wrote up the Cryostat path that actually matters: Quarkus on minikube, operator-driven agent injection, and JFR capture inside Kubernetes. Useful if you want a local lab before doing this for real. buff.ly/cTQltiA #Quarkus #Kubernetes

Phillip Kruger pointed me to a small Quarkus Agent MCP test thread, and it says more than most benchmark decks. My takeaway: skills earn their keep when they cut wrong turns, retries, and stale framework guesses. buff.ly/G2ck1iJ

I like build features that come with numbers. This piece walks through Quarkus JAR tree shaking on two small JVM services, shows the real size drop, and calls out the dynamic-loading edge where you need exclusions. buff.ly/SmWqQMc #Quarkus #Java

Happy birthday @andresalmiray.com 🎂🎉🎂🥳🎊

I get suspicious when trim().toUpperCase() turns into an agent. This walkthrough shows where MCP tools fit in a LangChain4j workflow: one Quarkus MCP server, one workflow app, /topology, and tests against Ollama. buff.ly/OiFml9q

AI did not kill the need for senior developers. It made judgment more valuable. The old brakes on bad ideas were time and bandwidth. Now the real job is deciding what should not exist. New piece: buff.ly/qWIKVwS

I do not trust “safe logging” until a test proves the token is gone. This piece shows how Quarkus REST client header masking actually works, including the easy footgun around replacing the default list. buff.ly/jxuTxit #Quarkus #Java

Static agent diagrams age fast. This piece shows how to expose a live LangChain4j workflow from Quarkus with topology HTML and an SSE run feed. buff.ly/BY4kbsx #Quarkus #LangChain4j

Agent names are not a routing strategy. This Quarkus + #LangChain4j walkthrough uses Skills to make ownership explicit, keeps a baseline path without them, and shows the routing difference with tests. buff.ly/9NGJR5N #Quarkus

50 tools later, assistant problems stop looking like prompt problems. I built a Quarkus + LangChain4j + Ollama example to show what tool search actually buys you: a smaller working set and more prompt headroom. buff.ly/zHa5ZPc #Quarkus

#JCON2026 keynote is live: 'Code Is Cheap. #Software Isn’t.' by yours truly. #AI can now write #code faster than most #developers can #review it. That sounds like progress. In many teams, it isn’t. In real #Java systems, the problem was… Grab your ☕️ and hit play: youtu.be/Di4ii6Xsb9A

My demo failed because Podman wasn’t running and Bob started reading an infrastructure problem like code. That turned into a better Quarkus Agent MCP. I wrote about what the tool actually changes in Quarkus work. buff.ly/l2G3SGV

In this #InfoQ #podcast, Andy Damevin explores #Roq - a static site generator built on top of #Quarkus. He covers: • The motivation behind the project • Why Java & Quarkus were chosen • Migration to Roq • Its future direction 🎧 Listen now: bit.ly/3QQZF5V #Java #SoftwareDevelopment

Sensible framing by IBM CEO Arvind Krishna at this #Ibmthink event. I mean - if you're going to make the case for industry transformation you might as well look back as well as forward when you're IBM. There is no point trying to look like a 2 year old AI startup. The historical photos...

AI adoption metrics are useful telemetry. They are terrible productivity metrics. The real question: did agents improve the SDLC, or did we just decorate the burn rate? buff.ly/lvRkJCp

When Bob can open the page, that is useful. When Bob can inspect console errors, failed network requests, screenshots, and performance traces from the same Chrome session I use myself, debugging gets much more grounded. New Main Thread post: buff.ly/6D4wlp4

comprehend Deep codebase understanding for AI coding agents, without smashing your context window.

The old tutorial contract was simple: follow the steps, get the same result I got. AI coding assistants break that. If every reader gets a different implementation, tutorials have to teach evaluation, not matching. buff.ly/rVTxQ1h

The more I use AI coding agents, the less I trust walls of "don't". Better prompts are boring: goal, context, boundaries, workflow, stop conditions. I wrote about that through a parenting lens, using IBM Bob as the case study. buff.ly/o1kUR9S

The more I use agents, the less I think "prompting" is the right frame. The real work is context transfer: getting the picture in your head across the gap before the model draws the wrong thing beautifully. buff.ly/UzRGTRf

A brief introduction to Quarkus Flow buff.ly/DOWZ095 #quarkus #cncf #workflow #java

Java modernization estimates should not stop at file counts. I wrote a spiderweb model for AI-assisted work: scope, semantics, knowledge opacity, integration/data, feedback friction, and governance. buff.ly/l9cimaO #Java #Modernization

Enterprise AI agents are not blocked by lack of information. They are blocked by trusted information. Frameworks, blueprints, best practices, and runbooks need owners, metadata, access checks, and evaluation. MCP is only the front door. buff.ly/rVAqIMe

We are pleased to announce the release of A2A Java SDK 1.0.0.Beta1, our first release fully aligned with the final A2A Specification 1.0.0. medium.com/@emmanuel.hu... #A2A #Java #Agents #AI

I have become suspicious of long agent command lists. For serious Java repos, I want repo-local skills that carry the team's rules with them: LSP navigation, ArchUnit, OpenRewrite, Snyk, Context7, flow tracing, and test gaps. buff.ly/bYf9gru

The more I use coding agents, the less I believe in "just let it figure it out." The hard part was always specification. Agents make it cheaper to execute and easier to dodge. I wrote about the spec trap and where the sweet spot moves. buff.ly/0XUXTXs

“I don’t think there’s a damn thing we can do to stop the train. But we can certainly control the culture, since the culture is us.” We are our own worst enemies

Quarkus cache invalidation skips failed methods by design. I turned the edge case into a small pricing-service tutorial with `@CacheResult`, `Uni`, `@CacheInvalidateAll`, and a `CacheManager` escape hatch. buff.ly/31VBDmL

I like #SBOM tooling most when it is boring and close to the build. This write-up walks through `quarkus-cyclonedx`: what lands in `target/quarkus-run-cyclonedx.json`, how that differs from `quarkus:dependency-sbom`, and how to validate both before CI archives them. buff.ly/mqbckoE #Quarkus

Building agent-ready RAG in Java without pretending enterprise PDFs are already nice Markdown. #Quarkus + #Docling + pgvector + #Ollama, with background indexing, readiness, retrieval logging, and guardrails. Small enough to run locally, shaped like something you can extend. buff.ly/JpI936E

I moved IBM Bob off my Mac and onto a remote Linux workspace over SSH, using Podman to simulate the remote host locally. The useful part was debugging the real issues, not just showing the final setup. Step-by-step write-up: buff.ly/nfGOj2g #Java #Podman #SSH

Don't miss these great Quarkus sessions today at JCON EUROPE... 3 Years of Quarkus in Production, What Have We Learned? Speaker: Jago de Vreede Time/Room: 09:00-09:45 at Cinema Black Box (1/4)

I keep seeing people say AI coding tools break down on “real” Java codebases. A lot of the time, the problem is simpler: the tool is guessing because we gave it no map, no conventions, and no real repo context. This piece shows how I’d fix that with JDTLS + MCP. buff.ly/SQ5EmN7 #Java #AI #MCP

Java is becoming a better scripting language without becoming a toy scripting language. JBang and modern Java make automation feel less disposable. buff.ly/cf1wqHa

JCON Chasing the Main Thread - Adventures in AI Assisted Coding

A digital badge is easy. A badge you can actually trust is a different thing. I wrote a hands-on Quarkus tutorial for building a credentialing platform with signed assertions, HMAC-secured webhooks, Flyway, PostgreSQL, and Qute. buff.ly/XSbx8tT #Java #Quarkus

I use AI every day, and I’m less convinced the real gain is productivity. For senior Java devs, a lot of the work just moved: less typing, more review, more judgment, more mental load. buff.ly/fyU31Vx

AI coding agents rarely fail because they do not know syntax. They fail because they do not know the system: rules, constraints, tests, architecture, and tradeoffs. buff.ly/3twSANo

Make sure you attend the Quarkus Workshop today at JCON EUROPE. "Think AI is only for Python? Think again. A hands-on Java AI workshop that will change your mind!" with Don Bourne & Michal Broz. github.com/quarkusio/qu... #quarkusworldtour

I use AI every day, and I’m less convinced the real gain is productivity. For senior Java devs, a lot of the work just moved: less typing, more review, more judgment, more mental load. buff.ly/fyU31Vx

Architecture does not drift because teams are careless. It drifts because boundaries are invisible until code, tests, and build rules make them visible. buff.ly/ZjhXHm5

Java is not replacing Python for every AI task. But once AI work becomes jobs, APIs, auth, observability, and deployment, Java starts looking very adult. buff.ly/f0BNKSD

I’ve been thinking a lot about how teams review agent system prompts. Many still write them like onboarding docs for humans. Friendly, vague, and full of assumptions. That is a problem once an autonomous coding agent starts editing real repos. buff.ly/PzB13Eh #AI #DevTools #Java #PromptEngineering

Ported skills.sh from TypeScript to Java+Quarkus in one AI-assisted session. 196 tests, ~10ms native binary, zero Node.js. https://maxandersen.github.io/jskills/ https://www.linkedin.com/posts/maxrydahlandersen_the-agent-skills-directory-activity-7451209487178297344-4JDT

Still true: AI coding tools reward compounding context, not bigger prompts. For senior Java devs, the win is smaller loops, sharper constraints, and better review habits. buff.ly/vOKgaXY

Code got fast. Software didn’t get easier. History already solved this problem. We just forgot to apply it. Toyota > AI hype. buff.ly/5Laryzu #AI #Java #SoftwareEngineering #Dev

Most AI agents are still just chatbots with tools. Real agents need scheduling, retries, persistence, and execution that survives restarts. Java already has strong answers for that. Guest post by Ronald Dehuysser: buff.ly/YWbxy2g #Java #AIAgents #JobRunr