AIM INTERVIEWS

Who are you?

My name is Jean-Baptiste Kempf. I'm known as the VLC guy. I'm a French engineer and startup creator. I've been the president of VideoLAN for about 15 years, and the lead developer of VLC for even longer. I also run several small companies around open source, like VideoLabs and FFlabs.

I'm now the founder of Kyber, a new startup building technology for real-time machine control, using the lowest-latency video possible.

At my core, I'm a humanist and someone who cares about engineering done right, software that lasts, and keeping the Internet fun and open.

You got involved with VLC in 2003 while still at university and became a core leader of the project. Why did you decide to get involved? Was there any inspiration?

Honestly, it wasn't a grand plan. I arrived at Ecole Centrale Paris in 2003 and became vice-president of the student network (the whole campus infrastructure was run by students), which was unusual even in France. VideoLAN was already there, built by earlier students who'd created it to stream TV across the campus network. I started helping out because I genuinely found the problem interesting: I was a movie and TV show fan and it was a cool way to improve my programming skills. I got more involved during two internships where I had a bit of free time. I ended up maintaining the Windows port because nobody else wanted to do it. That's how you make yourself useful in open source: do the things others avoid.

By 2015, VLC media player had over 2 billion downloads across the globe. The VideoLAN community you led built ports that supported virtually every desktop/mobile platform and video format. You also famously turned down numerous offers to sell the project. What was the most surprising thing that happened while running VLC? What did you learn?

The most surprising thing was probably just how far it went. When it was just a student project on a campus network, no one imagined it would be on a billion devices. The cone logo was a dumb joke that stuck -- that alone tells you something about how little we expected.

During the work on VLC, I had to do a ton of weird things, notably non-technical ones, on legal, community, personal matters, with unusual and amazing rencontres with magical people.

The reason VLC succeeded was precisely because we had no business agenda. No marketing, no monetization pressure, just: let's make something that works well and makes people happy. The moment you put revenue goals above engineering quality, you lose the thing that made users trust you. I turned down acquisition offers because those offers were not moral. They would have destroyed the project. The non-profit structure was deliberate: it meant the project would outlive any individual or company's interests, including mine.

What I learned from all this was what I wanted to do with my life: make complex things look simple and available to many; and, ultimately, be happy.

Your work on Kyber aims to make real-time video streaming and machine control nearly instantaneous (8-10ms). What new experiences or capabilities do you envision will be uniquely unlocked from this level of ultra-low latency? Why did you choose QUIC as the foundation, rather than conventional approaches like sockets or WebRTC?

We have great solutions for broadcast streaming and person-to-person communication, but real-time machine control is still wide open. And there will be more and more autonomous machines in the wild, that need control and supervision. Kyber is this.

QUIC is done for low-latency and avoiding head-of-line issues. And it is based on UDP. Traditional sockets don't fit for this need.

WebRTC is too complex to scale and opens too many sockets, while QUIC is a single-socket, TLS by default, and can mux multiple feeds in the same socket. Also, there is no difference between video and data channels, which makes WebRTC quite convoluted.

If you had unlimited resources (time, money, etc.), how would you spend your time?

I'd double down on the problems that are genuinely unsolved. Ultra-low latency is one of them.

Beyond that? I'd push for better open-source video intelligence and distributed content: tools that help people track what they're watching, discover what to watch next, know where to find it legally. The playback problem is largely solved; the discovery and access problem is not.

I'd also invest heavily in other open source solutions, from codecs to infrastructure, to other desktop and mobile open source projects. Doing things for many is very exciting.

What's a message you have for the world?

You can achieve great things. Believe in yourself.

You don't need a business plan, a big budget, or permission from anyone. A small group of students with a good problem and enough stubbornness can end up on a billion devices. So just start, do the things others avoid, and make something that makes you and people happy.