Luka Kladarić

Luka Kladaric is a cloud infrastructure and architecture specialist with twenty years of experience cleaning up other people’s systems as a consultant, architect, and the person called when production is on fire & nobody can explain why.

He’s currently CPTO and co-founder of rmBug, a database security startup. He writes about engineering sustainability and the things that break when you stop paying attention.

Talk: The Maintainer, the Model, and the Merge Queue

Open source won. Linux runs on everything from your phone to the ISS. Kubernetes is the default deployment target. PostgreSQL is eating Oracle’s lunch. The entire public cloud runs on open source software managed by people who mostly aren’t paid to do it. That victory happened, and then AI showed up with pull requests.

Three things are breaking at the same time.

The chardet licensing fight earlier this year was a preview. A Python library maintainer used Claude to rewrite the project from scratch, switched the license from LGPL to MIT, and a plagiarism detector came back at 0.04% similarity. Is it a derivative work? Nobody knows. No court has ruled. When one does, the precedent applies to every copyleft project in every language. Clean-room reimplementation assumed you’d either seen the code or you hadn’t. That distinction stopped making sense when the model trained on all of it.

On the maintainer side, I’ve been watching what happens to projects I use and occasionally contribute to — things like dnscontrol, where I’ve sent a couple of PRs. Even on smaller projects you can see the pattern forming: AI-generated submissions that look plausible and cost more time to review than they took to generate. Maintainers already stretched thin on triage are about to deal with a volume problem they have no tooling for.

And then there’s the contributor side, which is where it gets personal. I hadn’t shipped my own code in a decade. January this year I hit a missing feature in dnscontrol, built it with Claude Code open in the next terminal, and something clicked. Six small projects in three months. The activation energy dropped, I stopped deciding which ideas were worth building, and I just built them. That happened to me, and it’s going to happen to a lot of people. The question is what that does to a contribution model that relied on effort as a quality filter.

Licensing is broken. Triage is about to get harder. The contributor pool is about to get wider and shallower. This talk is about all three, and what open source governance needs to look like on the other side.

Get your free ticket here https://www.entrio.hr/en/event/dorscluc-2026-28348 and see you at DORS/CLUC 31!