Top Takeaways from Devoxx Belgium 2026

In October 2026, I visited Devoxx Belgium and again it was an awesome event! I learned a lot and received quite some information which I do not want to withhold from you. In this blog, you can find my takeaways of Devoxx Belgium 2026!

1. Introduction

Devoxx Belgium is the largest Java conference in Europe. This year, it was already the 23rd edition. As always, Devoxx is being held in the fantastic theatres of Kinepolis Antwerp. Each year, there is a rush on the tickets. Tickets are released in several batches, so if you could not have a ticket during the first batch, you will get another chance.

The first two days of Devoxx are Deep Dive days where you can enjoy more in-depth talks (about 2-3 hours) and hands-on workshops. Days three up and including five are the Conference Days where talks are being held in a time frame of about 40-50 minutes. You receive a lot of information!

This edition was again a special one for me, because I got the opportunity to speak at Devoxx myself, which has been an awesome experience! I gave a Tool-in-Action talk on Monday, but more on that later on. It was even more special, because this year I could also participate as a member of the Program Committee which gave my some insights in how a balanced conference program is created.

Enough for the introduction, the next paragraphs contain my takeaways from Devoxx. This only scratches the surface of a topic, but it should be enough in order to make you curious to dive a bit deeper into the topic yourself.

Do check out the Devoxx Youtube channel. All the sessions are recorded and can be viewed there. If you intend to view them all: there are over 200 of them…

The featured image is taken by the great event photographer Dimitris Doutsiopoulos.

2. Artificial Intelligence

Let’s start with AI first. More and more AI-related talks are given, even the non-AI tracks often are influenced by AI. We cannot neglect the importance of AI which transforms are way of working very rapidly.

2.1 Spec-Driven Development

I am personally very charmed by Spec-Driven Development (SDD) since it correlates very much to the way I am doing software engineering. In that regard, you must definitely take a look at the Deep Dive From Developer to Builder: A Spec-Driven Development Deep Dive by Simon Martinelli. The AI Unified Process is clearly explained here and you will get some good insights how to apply this yourself. It all starts with the specifications and I can really recommend this approach: I applied it successfully myself. A remarkable note to me was that for a particular client, no PR’s were executed anymore. Instead, trunk based development was used and afterwards some kind of pair review was executed. This is something I heard quite often at other AI related talks: reviewing starts becoming a bottleneck and you cannot review everything. Instead, ensure that you have enough deterministic checks in place (e.g. linters, cross-checks, static analysis, etc.).

2.2 Loop Engineering

Another hot topic in AI-assisted engineering is loop engineering. This term was first coined by Addy Osmani in his blog about loop engineering. There are different kinds of loops nowadays, very clearly explained in the Deep Dive Loop Engineering, beyond prompt, context, and harness engineering by Guillaume Laforge and Wietse Venema. Your loop needs a specific goal, it needs to know when to stop. Also, ensure that the goal is reachable (e.g. asking for a 100% test coverage will likely never be met). There are quite some problems to tackle when using loop engineering. However, I do think this is my next step in my AI learning path.

Do also take a look at the Software Engineering is Becoming Factory Engineering talk at the AI Engineer World’s Fair 2026 by Zach Lloyd. It will take you only 20 minutes, but worth watching.

2.3 Software Factory

And now that you are already watching Deep Dives, do check out Responsibly Dimming the Lights in the Software Factory by Victor Rentea. It is always a joy to watch Victor presenting, but it can be quite overwhelming to visit a Deep Dive of Victor. Some of my notes:

• You can use the grill-me skill to iterate on an idea.
• Use Spec-Driven Development in a loop, this relates to the previous Deep Dives.
• If you do not understand what the AI model wants to do or suggests, ask it to explain itself. This way, you will learn. Not only useful for junior engineers, but just for any engineer if you do not understand the response of the AI model.
• Before letting a human review specs, code, etc. first let AI review it. Only autofix the important ones and do not autofix or fix the nitpick issues.
• If AI generates a lot of code, and you want a human to review it, ask the AI to split it into multiple smaller PR’s.
• If you have enough confidence in the automated loop, including review cycle, you can start with reviewing less.

2.4 Sandboxes

Do you also run agents on your laptop? If you do, this is quite risky. Especially, when you are using skills, MCP servers, etc. The AI agent can do many things on your laptop, even harmful things. Docker has a solution for this, and that are sandboxes. It was one of the opening keynotes, A Fleet of AI agents, Each in Its Own Sandbox: Inside Docker’s Agentic Platform by Jean Laurent de Morlhon. A Docker sandbox is a microVM which contains the agent and a docker daemon. By using a DockerKit you can package software and configuration for the sandbox. It looks very promising and prevents that an agent can do all kinds of unwanted things on your laptop. Do check out Docker Sandboxes in order to learn more about it.

2.5 Groma.md

Groma.md is from the creator of Backlog.md, Alex Gravilescu. It scans your code and creates C4 diagrams from it. This way, you have a markdown representation of your architecture. The idea for Groma.md came because Alex ran a dark software factory and did not have any clue how the code was set up. Groma.md gives you these insights. I tried this on one of my projects, and it just worked. I do not know yet where it fits in the workflow, because I create my architecture first before writing any code. I talked about this with Alex and he will work on that somewhere in the future. There was a keynote, a Tools-in-Action and a conference talk about Groma.md.

2.6 Rise of Agent Enablement

The last one of this AI section is The Rise of Agent Enablement: Coding Agents Don’t Scale Themselves. Neither Do Your Teams by Patrick Debois (you know, the DevOps guy). Great insightful talk, some of notes:

• Stop building the code, build the thing that builds the product for you.
• If you want to keep doing the technical stuff, join the platform team which will build the platform. Otherwise, you will shift left (to the business side) or shift right (to the QA side).
• Apply skills, then loops, then build the software factory.
• Problems with software factories are concurrency (what if many agents change the same code?) and QA.
• Concurrent Delivery will become the challenge.
• Review is not the bottleneck, risk assessment is. If a change is low risk, use extra checking tools and auto-approve the change. Also with human reviews there was a risk. You could label code or changes with a particular risk and based on that, define what kind of review is needed.

3. Security

Security talks tend to get more and more popular nowadays. Is this also the influence of AI? Who knows, but it always had my attention and I want to mention two talks I visited.

3.1 EU Cyber Resilience Act

If you did not hear about the EU Cyber Resilience Act (CRA) yet, then you are quite late at the party. Since the 11th of September, you are obliged to notify your customers about specific security incidents. From the 11th of December of 2027, most of the other CRA rules will become mandatory. In CRAcking Compliance – The EU Cyber Resilience Act by Christian Schabetsberger, you get a good overview of what it means to you. If you already have set up generating and scanning SBOM’s, you are already on the good path. But that is not everything you need to do.

3.2 Toxic Flows in MCP Servers and Skills

In the talk Toxic Flows in MCP Servers and Skills: What Your Agent Blindly Trusts by Brian Vermeer, Brian demonstrates how MCP Servers and Skills can invoke unwanted behavior. Definitely worth watching if you are using MCP Servers and Skills. Snyk, the company he works for, also open sourced an agent scan tool which you can use to scan your MCP Servers and Skills.

4. Other

In this last section, an overview of some interesting talks which do not relate to AI or security.

4.1 Spring Boot Testing

Philip Riecks shares for many years interesting content about Spring Boot Testing on his website. In this Deep Dive Prompt It Right: Spring Boot Testing in the AI Era, he gives in the first part an introduction into Spring Boot Testing, and in the second part how to apply this in the AI era. An interesting tool in order to use when you have slow tests, is the Spring Test Profiler. This tool gives you insights in how the Spring Context Caching is being used (or not used).

4.2 BootUI

An interesting open source component when developing Spring Boot or Quarkus applications, is BootUI. It is a local-only developer console which gives you all kinds of insights into your application. Do check out BootUI, the missing developer console for Spring Boot and Quarkus by Julien Dubois.

4.3 AsyncAPI

I am using AsyncAPI (and OpenAPI) for a while now. With OpenAPI you have a nice maven plugin in order to generate code based on the OpenAPI spec. I did not know how to do this with AsyncAPI. No Maven plugin exists, but in this talk AsyncAPI in action: OpenAPI for event driven applications, Michel Schudel shows you how to do so by means of the asyncapi CLI command. With the help of JavaScript templates, you can generate your code. Interesting and practical Tools-in-Action talk. Examples can be found in this GitHub repository.

4.4 Java

Devoxx is a Java conference, we cannot ignore the improvements that are being made to the Java language itself. Nicolai Parlog gives a nice overview in Java Next – From Valhalla to Leyden, from Babylon to Panama, from Amber to Detroit. If you do not know what these names are all about, I added a small overview of what these projects mean:

• Project Amber: Makes Java easier to write and read by introducing language improvements.
• Project Leyden: Improves Java application startup time, warm-up, and footprint.
• Project Valhalla: Explores value classes and primitive classes to enable more memory-efficient data representations.
• Project Babylon: Extends Java’s ability to represent and analyze code, including code reflection and interoperability with accelerators such as GPUs.
• Project Loom: Introduces lightweight virtual threads and structured concurrency to make highly concurrent applications easier to develop.
• Project Panama: Improves interoperability between Java and native code, including foreign-memory access and foreign-function calls without requiring JNI for many use cases.
• Project Lilliput: Reduces the size of Java object headers to lower memory consumption and potentially improve application performance.
• Project Detroit: Is an OpenJDK initiative that provides native implementations of the standard javax.script API to allow seamless interoperability between Java and external scripting languages, specifically JavaScript and Python.

4.5 Testing Architecture Made Simple

To conclude with, I should not forget about myself. I had the honor to present a Tools-in-Action about the Taikai library. Again, it was awesome to stand in front of a huge cinema room and share knowledge. If you are interested in getting up to speed with ArchUnit in 5 minutes and write maintainable ArchUnit tests, do check out my talk Testing Architecture Made Simple.

5. Conclusion

Devoxx 2026 was great and I am glad I was able to attend the event. As you can read in this blog, I learned a lot and I need to take a closer look at many topics. At least I do not need to search for inspiration for future blogs!