A new chapter
After sixteen years at Collabora, I’m moving on: stepping down as maintainer of WirePlumber and starting a business focused on software engineering in the age of AI.
Exactly 20 years ago, I entered university. I was enrolled in the Computer Science Department of the University of Crete. The subject was so fascinating that I didn’t want to miss any class, except for all the math classes that I found boring at the time (even though I was really good at math in high school). The problem with math, as I realized later, was that it wasn’t very creative. In computer science classes, we were creating pieces of software and connecting cables on breadboards to make computers actually do something interesting. Math wasn’t anywhere near that level of fun!
Just a few years later, I lost interest in classes. Being a Linux user since before university, I soon realized that I was in a position to contribute to software that I was using. This was far more interesting and creative and it was actually improving my own user experience, a great motivation on its own. Not long after, other people in the community started recognizing my contributions and my skills. I felt great about it and it only kept fueling me more, giving me a sense of purpose: helping other people through my contributions.
In 2009, I was encouraged to apply to the Google Summer of Code program, as a student. I hadn’t really given it much thought myself, but it sounded like a good idea: I was going to be paid to write software that would be useful for the community. What’s not to love about that? It fed my creativity and my sense of purpose, and it supported me financially at the same time.
This GSoC project opened another chapter in my career, as it led to me being employed by Collabora just about a year later, in 2010, during its initial growth hiring era. I still hadn’t graduated, though. And the longer I stayed at Collabora, the less I wanted to graduate. For the same reason that math had lost to computer science and classes had lost to open-source contributions, the degree was now losing to the job it was supposed to lead to. I had long outgrown what I was being taught at university. I didn’t want to leave unfinished business behind, though, so I eventually graduated in 2013, working part-time and attending classes in the meantime. But my heart had already moved on.
Sixteen years, and very little of it was boring. Every now and then there was a new customer, a new domain, a codebase to learn from scratch and a different problem to solve. Through Collabora, I was able to see my code being shipped on millions of devices across multiple industries: in-vehicle and aerospace infotainment, household entertainment, digital signage, security cameras, healthcare, and many more. And beyond just watching, I was on the frontline helping to shape the technology of countless such devices. All while staying true to the principle: use open source, contribute everything back to the community and most importantly influence the direction of the upstream projects to ensure alignment with real-world product needs. I got to learn, to grow, and to make a living doing something I loved. I was truly lucky.
All good things come to an end, though. After all these years, I have decided to move on. My last day at Collabora was Friday, 11 September 2026. I want to thank my colleagues for their trust in me over the years and for having given me the opportunity to become the professional that I am today. Sixteen years is a long time, and this was not a decision I made lightly.
Why?
Because I’m restless. I am constantly looking for a new challenge and most importantly, I am eager to create. Creativity is, unfortunately, a part of me that I neglected over the years in this job. Partly because in consulting there aren’t many opportunities to create new things, but partly also because the stress of everyday obligations didn’t leave me much energy to bring out my creative self.
We all change as we grow up, and I am no exception. My life goals have changed and my priorities have shifted. Or perhaps I shouldn’t say “changed”, but “shaped up”. I am not a different person; I am just clearer than ever about who I really am.
Looking back, there’s a pattern. From my first open-source contributions to the products I helped shape at Collabora, what has always driven me is my eagerness to help others by creating things that solve their problems. But consulting always came with the constraint of working within someone else’s vision. What I have realized is that I need the freedom to express my own ideas, on my own terms. This is the path of entrepreneurship, plain and simple. And it’s time to follow it.
What’s next?
Over the past few years, I’ve been building with AI models and agents. I find them fascinating. You come up with a software idea, you steer the agent to implement it, and in a very short time you have something in the flesh. I may have been hesitant at first, but nowadays I can’t imagine myself writing software without them.
When I was younger, in my university years, I used to have lots of software ideas. I had real problems that I wanted to solve for myself and for the people around me. And I solved some of them… I remember first creating a small timer app for my younger brothers, to track how much time they were spending playing games on the computer. Then it was my first open-source app, QtWine, a GUI for Wine to help me manage all the Windows software that I was still using after migrating to Linux (because how could I live without Windows Media Player? – don’t judge me, I was practically still a kid). Then my violin teacher asked me to help him track his student records and I created yet another app to do that. Even though that could have easily been done in a spreadsheet or Microsoft Access, a custom app had a much nicer user experience; and besides, it felt good to create something new.
As the years passed by, I started having bigger problems to solve and (much) more ambitious ideas. I tried to solve some of them, but reality kicked in very fast. And the reality was that writing software required a substantial investment of time and energy, which I had very little of, thanks to working a full-time job. This went even further: after years of being a hardcore Linux and open-source hacker, I started using proprietary software, migrated to macOS and compromised in many ways, just to try and save as much time and energy as I could, so I could funnel it into the things that matter to me.
Times are changing, though. Large Language Models (LLMs) are making it easier and cheaper than ever to create software. It no longer takes a substantial investment of time and energy. You can make something completely custom, behaving exactly the way you want it to, looking exactly how you like it, in a matter of days. There’s no reason to compromise anymore… nearly anything is possible. It’s a dream come true!
But I hear you. You are already thinking… “easy”? and “cheap”? Of course it’s not that simple at all. The price hasn’t disappeared; it has moved. It’s easy and cheap to express ideas, but making a real-world product that needs to be secured and maintained for the long term is still a challenge. And this is where software engineering experience matters.
What I’m working on
So, what am I going to do about this? I am starting a business. The specifics are still taking shape and I’d rather not go into them just yet, but I can tell you where my focus is: software engineering in the age of AI.
Agents can feel magical, but if you look under the hood, they are far from it. An agent starts every session from nothing. It doesn’t remember yesterday, it doesn’t know why the code looks the way it does and, if you let it, it will confidently walk in the wrong direction. Mistakes are cheap to make and expensive to discover. Keeping it on track is pure engineering. You need to decide what the agent gets to see, how to split the problem into small pieces of work and which models to use for which task. You need a testing strategy, and a codebase organized well enough for the agent to find its way around.
The other side of the same coin is putting that intelligence inside the software itself. We are already seeing applications that you can simply talk to, or ones with a model quietly assisting you in the background, handling what can’t be expressed as a fixed set of rules. I believe that we are going to see much more of that in the future. Not necessarily a chatbot bolted onto everything, but intelligence where it actually improves the user experience. And here too, the model is a component that forgets, that can be confidently wrong and that costs money on every request. Making it dependable is an engineering problem, not a magic trick.
This is where my background comes in. For sixteen years, I’ve contributed to and maintained open-source projects, reviewing code I didn’t write and keeping it coherent while the project grew, and I’ve learned what it takes to keep a project in shape. Humans actually need the same things that an agent needs: an organized codebase where they can find their way around, the chance to focus on one part of it at a time, small iterations, and a testing strategy. This is even more evident in open-source projects, where a contributor may be starting from scratch, without any prior knowledge of the codebase and without any onboarding, just like an AI agent. And as a consultant, I’ve been that newcomer more times than I can count, in both open-source and proprietary customer codebases.
Moreover, I’ve integrated open-source components into products running on millions of devices, where mistakes can be expensive. A model inside an application is just another such component, only a less predictable one. Keeping things working there takes discipline. The tools and the problems may be new, but the engineering around them is the same.
Let’s talk
Since things are still taking shape, this is the best time to talk. I’d love to hear what you think about all this, whether you love it or hate it. And most of all, I’d love to hear about your problems: where AI agents are letting you down, what it is that you’d like to optimize, or where you wish your software were a little smarter.
You can reach me at [email protected], or find me on LinkedIn, Threads or Bluesky.
If you’d rather just follow along, I’ll be writing on this blog about how it goes. There’s an RSS feed that you can add to your feed reader. I also intend to revamp this website soon, to reflect what I am doing now.
What about WirePlumber and the other projects?
I am stepping down as the maintainer of WirePlumber and my long-time colleague Julian Bouzas is taking over. Julian has been working on WirePlumber next to me since the beginning of the project in 2019. He knows the codebase as well as I do, and he has been working on it lately much more than I have. He is the person who has been making the last few releases as well and will keep doing that in the future. I trust that the project is in good hands.
I am not vanishing overnight, though. I’ll be assisting Julian with the upcoming 0.6 release in the next few weeks, which is mostly about paying off technical debt: reworking and simplifying bits and pieces to fit the session management use cases better. I also recently triaged a lot of tickets that were sitting open long past their relevance. I’d rather hand the project over in a tidier state than the one it’s in today.
More broadly, I don’t foresee continuing to contribute to the projects I have been involved with: WirePlumber, PipeWire, GStreamer, BlueZ, AGL. Not because I don’t like them, but because there are only so many hours in a day, and the ones I have now belong to something new.
That doesn’t mean disappearing. I’ll still be around on the socials, reachable by mail and by tagging me on GitLab/GitHub. I’m happy to answer questions about why something was designed the way it was or participate in discussions about future directions. The full context around why certain choices have been made in the past is the one thing that unfortunately doesn’t live in git.
Sixteen years of my work is out there in the open, running on millions of devices, and it will keep running without me, in the hands of people who care about it as much as I did. Thank you to everyone I built it with.