Mind-Controlling a Drone: What CereBionics Is Building

Kitty Mayo watched Agnessa Pedersen control a drone in real time — not with a joystick, not with a phone, but with her mind. She called it one of the most moving moments of her career.

That's a strong thing to say. And I believe it.

These demos have existed in labs for years. You've seen the clips — a paralyzed patient moves a cursor, a monkey controls a robotic arm. Impressive. But always enclosed. Always a proof-of-concept in a controlled setting, surrounded by researchers and equipment.

This felt different. A drone, physical space, real time. The gap between the lab and the world got a little smaller.

Agnessa is the founder of CereBionics. The company builds non-invasive brain-computer interfaces — no surgery, no implants — designed specifically for robotic platforms.

The goal is to remove the physical controller entirely. No joystick. No touchscreen. No physical interface between your intention and the machine's action.

Non-invasive is the key word here. Most of the serious BCI research — including the work coming out of BrainGate — has relied on implanted electrodes. That produces cleaner signals. But it also requires a neurosurgeon and a patient willing to have hardware placed inside their skull.

CereBionics is betting that you can get far enough with external sensors to make it practical. Practical meaning: deployable, not just demonstrable.

Brain-computer interfaces as a concept are old. Decades old. The science has been there, at least in outline, for a long time.

The problem is deployment. Getting it out of the lab and into physical environments, running on real hardware, doing real tasks. That's what hasn't happened at scale.

It's a pattern you see in a lot of deep tech. The concept is proven early, then sits in research for twenty years while someone figures out how to make it actually work in the world. Battery tech. Voice recognition. Computer vision. The idea wasn't the hard part. The hard part was making it robust enough to hand to someone who isn't a researcher.

That's the wall CereBionics is trying to break through.

Here's the part of Agnessa's thinking that I find most interesting.

She doesn't see the future as thousands of specialized interfaces — one for your drone, one for your wheelchair, one for your robotic arm, one for your car. She sees one interface, or maybe a handful, general enough to talk to all of them.

That's a big claim. But think about what we already did with the touchscreen. A single interaction model — tap, swipe, pinch — that works across a phone, a tablet, an ATM, a car dashboard. We collapsed dozens of physical interfaces into one.

If a non-invasive BCI can do that for human-machine interaction broadly, the implications go well past drones. Nita Farahany's book Battle for Your Brain covers the legal and ethical territory this opens up. Worth reading alongside the technical side.

There's a reason the invasive vs. non-invasive distinction matters commercially, not just medically.

An implant requires a patient who has no other option — someone with ALS, a severe spinal injury, locked-in syndrome. The market is real, but it's small, and the regulatory path is long.

A non-invasive device is a consumer product. Or at least it could be. You put on a headset. You use it. You take it off. No hospital, no recovery time, no surgical risk.

That's a completely different business. Different scale, different distribution, different speed to market. It also means the interface has to work well enough that an ordinary person — not a trained patient in a clinical trial — can use it without frustration.

That's the harder engineering problem. And it's the one Agnessa and her team are working on.

A drone is a good demo object. It's visual, it moves in three dimensions, and it responds fast enough that you can see the lag — or the lack of it.

But the drone isn't the point. The point is that a person issued a command from inside their head and a machine in the physical world obeyed it. Without touching anything.

If that gets reliable — not lab-reliable, but deploy-it-anywhere reliable — then a lot of the physical scaffolding we've built around human-machine interaction starts to look unnecessary. Controllers, keyboards, switches, levers. All of it is a workaround for the fact that machines couldn't listen to what we actually wanted.

We're early. Very early. But Agnessa flew a drone with her mind, and Kitty Mayo cried a little. That's not nothing.