Six billion downloads. That is the footprint of VLC Media Player, the orange traffic-cone icon that became the universal language of digital video. Its lead developer, Jean-Baptiste Kempf, spent years perfecting the art of moving data across networks without a stutter. Now, he is applying that same obsession to a much more volatile medium: the physical world.

Kempf is betting that the future belongs to millions of autonomous drones and robots roaming our streets. To make that vision viable, he founded Kyber. It is an infrastructure layer built to synchronize video, audio, and sensor data in real time. If you are controlling a machine from across the globe, every millisecond counts. Kyber aims to make that lag disappear.

Why Physical AI Needs a New Backbone

The rise of physical AI has created a bottleneck. You can have the most sophisticated neural network in the world, but it is useless if the connection to the hardware is sluggish. "Physical AI is only as good as the underlying systems running it," the team at Lightspeed Venture Partners noted recently. They led Kyber’s $5 million seed round, signaling a belief that the plumbing of robotics is as valuable as the AI itself.

Kempf’s approach is rooted in his past. He built the early iteration of Kyber as a side project while serving as CTO at the cloud gaming firm Shadow. He realized that the challenges of streaming high-fidelity video to a gamer are remarkably similar to those of streaming telemetry to a remote drone operator. Both require extreme precision. Both hate latency.

Scaling Beyond the Handful

Most robotics companies today are playing in the shallow end. They manage fleets of a few thousand units, often using custom, proprietary software built in-house. Kempf argues that this model breaks the moment you try to scale to millions. Managing a fleet of ten is a hobby. Managing a fleet of ten million is a systemic engineering crisis.

Kyber is designed for that jump in scale. It handles the observability required to know if a machine is actually working, even when no human is watching. It allows for remote updates, eliminating the need to physically touch a device to fix a bug. It is a utility, not a toy.

The Open Source Playbook

True to his roots, Kempf is keeping the core of Kyber open source. It is a classic strategy for building a standard. By giving developers the tools to build on top of his infrastructure, he ensures that Kyber becomes the default choice for the next generation of robotics engineers.

But the company is not just a software project. It employs forward-deployed engineers—FDEs—who work directly with clients in defense, telecommunications, and industrial robotics. They are building the version of the software that everyone else can use, effectively commoditizing what used to be a bespoke, multi-million dollar problem.

Key Takeaways

  • Latency is the enemy: Kyber uses streaming technology to synchronize sensor data and control inputs, aiming to eliminate the lag that makes remote robotics dangerous or inefficient.
  • Scale matters: While current robotics fleets are small, Kyber is built to manage millions of devices, a threshold where current proprietary solutions typically fail.
  • The hybrid model: Kyber follows an open-source core model, supplemented by enterprise-grade productization and hands-on engineering support for high-stakes industries like defense and telco.

What This Means for Robotics

For the industry, this is a shift toward standardization. If Kyber succeeds, it will become the "VLC of robotics"—the invisible layer that makes complex, real-time remote operation feel mundane. The company is currently prioritizing three segments: robotics, drones, and remote IT access.

Kempf is not just building a tool. He is building a bridge. Whether that bridge holds up under the weight of millions of robots remains to be seen. The next eighteen months will be the test. If he can replicate his VLC success, the streets of the future might just run a little smoother.