This week I attended the Jfokus software development conference in Stockholm, Sweden. I first went in 2011, and I have been back many times through the years. The conference has a Java focus (duh!), but many talks cover general topics as well.
The whole development team at NGM got tickets. It is really nice to be able to discuss and compare notes with your colleagues. The big theme this year, apart from Java, was of course AI and LLMs.
Talks I Liked
The First 80% of Reading One Billion Rows Fast Enough by René Schwietzke. This is a talk on Java optimization, and I really enjoyed it! I had not heard about the challenge before. The input is one billion rows of simple weather csv data, and the idea is to see how fast it can be processed in Java. Of course there are solutions that use incredibly weird and obscure techniques, but this talk goes through some basic optimizations that together add up to a run time of 20% of the original solution.
After setting up the problem, René shows a base-line solution, and gives its run time. Then he goes through a number of optimizations, and shows how much each saves. Examples of optimizations are: replace split() by indexOf(), use int instead of double (we know there is only ever one decimal digit), mutate existing objects instead of creating new ones, only read bytes (not Strings, doubles etc) and delay Unicode processing, simpler Min/Max, create the hash code while traversing the line.
René used a flamegraph from a profiler to guide what areas of the program should be optimized. Some general rules he followed were: replace standard JDK functionality (it may be more general than what is needed), low or no memory allocation, avoid wrappers and immutability, reduce branching (if, loops), exploit what you know about the input data.
The Future of Work by Henrik Kniberg. This was second of two keynote talks that opened the conference. Henrik did a live demo where he used several AI agents to accomplish tasks, such as making code changes, creating a branch in git, and creating a PR with the changes. The AI agents have instructions in the form of short text documents, and these instructions can be updated, even by the agents themselves (subject to human approval).
The agents appear as their own users in Slack. They can also be given recurring tasks, for example to create a report each morning on a given subject, and to mail out the report and post a summary of it in Slack. He also demonstrated how they can trouble shoot if something goes wrong, for example if it can’t create a git branch. The LLM used in the demo was Claude, and Henrik used it in voice-input mode.
He ended the presentation with some reflections on the implications of this way of working. He, like me, has always loved programming. Will this way of working, with agents writing a large chunk of the code, or all code, mean the end of software development for humans? First he noted that this is similar to moving from punch cards, to assembler, and to compiled higher level languages. You move up one abstraction level. He also noted that what he liked about programming was making and creating things, not necessarily writing the actual code.
A very good and thought-provoking talk. Actually seeing agents in action, instead of merely being told what they can do, really brings the points home.
Ask the Architect with Brian Goetz (Java Language Architect) and Mark Reinhold (Chief Architect), both at Oracle. This was a Q&A session, where the audience had a chance to ask Brian and Mark Java questions. I didn’t really know what to expect from this session, since it will depend a lot on the questions asked. But I was pleasantly surprised. There was quite a variety of questions.
There were questions about serialization, GraalVM vs HotSpot, records, streams, Lombok, different deprecation modes, and more. Quite interesting. Before we started, Brian and Mark cautioned us not to begin questions with “Why don’t you just…”. Modifying a language that has been around for so long, with so many existing programs, is not easy. This became very clear when hearing some of the answers.
The Value of Conferences
Going to a conference is different from watching talks on YouTube, or reading books or blog posts about software development. It is nice to meet and talk to other developers. My standard question when chatting with other attendees is “What is your favorite talk so far, and why?”. At Jfokus, your name and your company are printed on your badge. This gives you another set of good icebreaker questions: “What do you do at Company X? What does the company do? What tech stack do you use?”. Also, when listening to a talk live, you have a chance to ask questions, either during the talk, or afterwards.
It is very convenient to attend a conference in your home city or country. It is cheaper and there is less travel. From the badges I saw at Jfokus, most people were from Sweden, and a few from Germany.
At Jfokus, there are usually six parallel talks, so it can sometimes be hard to pick what to listen to. When attending a conference, and listening to many talks in a row, you notice things that are mentioned more than once. Examples this year were LLMs checking the output of other LLMs, RAG (Retrieval-Augmented Generation), AI-assisted coding (with IDE plugins), GraalVM, and the Quarkus framework. After a conference I always end up with a long list of things to look up: techniques and tools I hadn’t heard about before, and books and articles to look into.
I also like to see which companies have exhibition booths at the conference. Even if I am not interested in their exact service, it gives me a general sense of what is popular right now. All the exhibitors get a little gold star in my book for sponsoring a conference.
Finally, it is always inspiring to go to a conference. Meeting people and learning about new ideas is exciting.
Odds and Ends
- Tobias Modig, in his talk The Developer Rhapsody, talked about “the mediocre developer”. Someone competent that stays at the same company for 15 years. The contrast is the brilliant developer, that gets bored after a year and a half and moves on. Who is more valuable for a company?
- “We think in generalities, but we live in detail” – Alfred North Whitehead. One of many good quotes Kevlin Henney mentioned in his talk Keeping It Simple.
- The venue, Stockholm Waterfront Congress Centre, is great. Everything worked smoothly, the food and the “fika” was great, and it is very easy to get to.
- The Jfokus web site could be improved. First of all, it would be good if the back button in the browser worked. Going back after clicking into a talk, you lose where you were in the schedule. Also, the link for rating the talk would be good to have next to the talk description.
Conclusion
Another great conference, with a good variety of talks and speakers. If you haven’t been to a conference in a while, try to find one to attend. It is really inspiring, and a nice way of keeping up with new ideas and technology.
Top comments (0)