Ljubiša Punoševac - JCrete® 2026 LinkedIn Post
4 August 2026
Just back from JCrete®. No CFP, no schedule, no slides. Every morning, you write a topic on a post-it, get twenty seconds to convince people it is worth their time, and then you find out whether anyone shows up. The agenda builds itself as the week goes on. I pitched three topics.
The first: if AI made us all faster, why are we not shipping software any faster? The room converged quickly. Writing code got faster, but writing code was never the bottleneck. The hard part is understanding an existing codebase and fitting new code into a system that already has opinions, and that part does not scale with AI.
The second was one concrete attempt at an answer from my side, running only the tests that can actually see your source code change. I have been working on per-test coverage mapping combined with diff analysis to cut CI feedback time. Nobody tried to talk me out of the approach, which I did not expect. What I got instead were the harder questions: whether JaCoCo is too heavy for this and going straight to ASM would be cleaner, and how any of it holds up under parallel execution and parameterized tests. I left with a longer list than I arrived with.
The third was about the cognitive load that comes with all of this. The tools got faster, expectations rose accordingly, and what is left for us is the part that never stops: reviewing, deciding, and being accountable. It was the last session on the last day, the slot where people usually start packing, and the room was full. We ran well past the scheduled end because nobody wanted to stop talking. That answered the question better than the discussion did.
What makes this format work is that nobody is performing. Small groups, experienced people, real disagreement, and enough time to actually finish a thought. Add good food, a lot of laughing, and Crete in July. Thanks to Heinz Kabutz and the disorganizers for making it happen.
#JCrete #Java
