The Infrastructure of Autonomy: How the Architect of VLC Is Building the Nervous System for the Robotics Age

If you have ever opened a video file on a computer, you have likely encountered the orange traffic-cone icon of VLC Media Player. With over 6 billion downloads, it stands as one of the most successful pieces of open-source software in history. But for its lead developer, Jean-Baptiste Kempf, VLC was merely a warm-up act for a much more complex challenge: the impending explosion of the physical AI economy.

Kempf, a French serial entrepreneur and a titan of the open-source community, believes we are on the precipice of a new era. In just a few years, he predicts that hundreds of millions of robots and autonomous drones will navigate our streets, factories, and skies. To manage this swarm, he is building Kyber, a Paris-based startup designed to serve as the critical infrastructure layer for real-time remote device control.

The Core Problem: Closing the Latency Gap

The premise of Kyber is deceptively simple: the physical world is messy, and remote control is notoriously laggy. When an operator or an AI agent sits thousands of miles away from a device, the time it takes to send a command, receive sensor data, and see the result—the "round trip"—must be virtually instantaneous.

"If you control things in the real world, every millisecond matters," Kempf explains.

Kyber acts as a high-performance SDK that synchronizes video, audio, sensor data, and control inputs across massive, distributed networks with minimal latency. By leveraging the same streaming technologies that made VLC a global standard, Kyber creates a high-fidelity "bridge" between the compute layer (where the decisions are made) and the action layer (where the machine moves).

Chronology: From Cloud Gaming to Physical AI

The genesis of Kyber can be traced back to Kempf’s tenure as CTO at Shadow, the cloud-gaming startup. At Shadow, Kempf was tasked with solving the "holy grail" of remote interaction: streaming a high-resolution, high-frame-rate video game to a user over the internet without the player perceiving any delay.

While that work was focused on consumer entertainment, Kempf realized that the underlying technology—optimized streaming, low-latency transmission, and edge-computing orchestration—could be repurposed for a far more vital application: industrial and autonomous robotics.

Following his time at Shadow, Kempf launched Kyber to address the "middle mile" of robotics. While proprietary, closed-source solutions exist for remote vehicle operation, they are often brittle, siloed, and incapable of scaling to the millions of devices Kempf envisions.

A $5 Million Vote of Confidence

The startup’s potential has already caught the eye of heavy hitters in the venture capital world. Lightspeed Venture Partners, the firm behind AI unicorns like Anthropic and Mistral AI, recently led a $5 million seed round in Kyber.

In a statement regarding the investment, the VC firm emphasized the limitations of current hardware-software integration: "Physical AI is only as good as the underlying systems running it."

This funding marks a significant validation of the "Physical AI" thesis—the idea that the current boom in large language models and neural networks will eventually shift from digital assistants to physical agents that interact with the real world. For Lightspeed, Kyber is the "plumbing" that will make that transition possible.

The Technical Edge: Why Scale Changes Everything

One of the most pressing questions in the robotics industry is how companies will manage fleets of thousands, or even millions, of devices.

"The largest fleets today have maybe 2,000 or 3,000 vehicles," Kempf notes. "Imagine you need to manage millions of them; that’s not the same thing."

Managing a handful of drones in a controlled environment is a software engineering problem; managing a global fleet of millions of robots is a systems architecture challenge. Kyber differentiates itself through three core capabilities:

1. Adaptive Streaming: Much like how VLC adapts to a user’s network speed, Kyber dynamically adjusts its data stream based on the available bandwidth and compute power of the remote device, ensuring that the control link is never severed.
2. Observability at Scale: As AI agents begin to replace human operators, the need for "machine-to-machine" observability becomes paramount. Kyber provides the telemetry required to know, in real-time, that a system is healthy, even if it is operating on the other side of the planet.
3. Remote Maintenance: By enabling secure, low-latency access, Kyber eliminates the need for technicians to physically travel to a device to push a software update or diagnose a bug.

Implications: The "Citrix" of the Physical World

While Kyber is being built with robotics and drones in mind, the platform has broad implications for remote IT access. Kempf envisions Kyber as a successor to legacy remote-access tools like Citrix or TeamViewer, but built for the modern, distributed, and AI-driven era.

The company is already in commercial deployment with customers across defense, telecommunications, and industrial automation. By offering an open-source core, Kempf is following the same playbook he used with VLC: build a tool that is so useful and accessible that it becomes the industry standard, then offer a productized, enterprise-grade version for large-scale, mission-critical deployments.

Furthermore, Kyber is bucking the trend of "software-only" startups by deploying Forward-Deployed Engineers (FDEs). These are technical experts who work directly with clients to integrate the SDK into complex, bespoke hardware environments. This model, reminiscent of the success of Palantir, ensures that the platform is not just a theoretical tool, but a reliable backbone for high-stakes industries.

The Road Ahead: A Global Ecosystem

With 25 full-time employees and offices in Paris, San Francisco, and Singapore, Kyber is positioning itself as a global infrastructure provider. The company is intentionally keeping its scope broad—targeting robotics, drones, and remote IT—to capture as much of the "remote-action" market as possible.

The philosophy behind the company is captured on its careers page: "The companies that tried to solve it spent years and tens of millions building custom solutions they’ll never share. We’re building the version everyone else can use."

As we look toward a future defined by autonomous agents, the bottleneck will not necessarily be the intelligence of the AI, but the speed and reliability of the nervous system that connects that intelligence to the physical world. If Jean-Baptiste Kempf has his way, the same orange icon that once revolutionized how we watch movies will be the foundation upon which the world’s robotic fleets are controlled.

In the transition from a digital-first to a physical-AI-first world, the companies that provide the "connective tissue" will be the most valuable players in the tech ecosystem. Kyber is not just betting on robots; it is betting on the necessity of the infrastructure required to keep them moving.