A patient fasts overnight, drinks two glasses of water and swallows a capsule.
Somewhere else entirely, a physician picks up a controller.
On the screen is a live view from inside that stomach, and the view is moving on purpose.
It banks left, holds against a fold, and pushes forward.
Ten minutes later the appointment is over and the hardware leaves on its own schedule.
So how do you drive anything through a stomach?
Why swimming is different at this size
Scale changes the rules of fluid.
A submarine coasts because its mass carries it forward after the propeller stops. Something a few millimeters long has almost no mass to carry.
At that size the ratio between inertia and stickiness collapses, and water starts behaving like honey.
Stop thrusting and you stop, instantly, in the same body length.
Worse, a propeller shaped like a boat propeller does almost nothing down there, because it is designed to push against a fluid that resists being shoved rather than one that simply clings.
So the propellers had to be redesigned, and so did the chassis around them.
What came out is not one screw but six directional jets: up, down, forward, back, left and right, driven by three motors, with an accelerometer on board telling the thing which way is which.
What is actually inside the shell
The body is injection molded polycarbonate, thinner than a sheet of paper in places, and it has to survive being bitten by someone who forgets to swallow.
The bite test runs to roughly 130 pounds of pressure.
At the nose sits a camera sensor ringed by four small lights, because the inside of an organ has no illumination of its own.
The motors driving the jets started life as the vibration motors in cell phones, about a sixth of an inch across.
Radio is the hard part.
A transmitter that small pushing a signal out through a body has almost nothing to work with, so a programmable chip on board compresses the video aggressively before it ever reaches the antenna.
The link runs at 915 megahertz to a dongle on a laptop, and the whole thing is single use.
How far along it actually is
The company behind it was incorporated in Hayward, California in March of 2019, and the work has run more than five years.
Early testing was as unglamorous as it sounds: the team swallowed more than twenty prototypes themselves, ran units in a cadaver lab, and flew buoyancy tests in a small aircraft to get a few seconds of weightlessness at a time.
Human trials began in New Zealand in 2024.
Investors have put in about $21 million since the start, including a $12 million round announced in January of 2025.
Control today is a game controller, with a phone app intended to replace it.
An early prototype produced 48 pixels of grayscale at a few frames a second, which is the honest measure of how far the imaging has had to come.
What it still cannot do
It is not approved, and the company says so on its own front page.
The device has not been cleared by the American regulator or any other, and it is not for sale or clinical use.
That is later than planned: clearance was once expected in 2025 with a launch this year, and the founder handed the chief executive job to an intellectual property lawyer in early 2025 and moved to advanced research.
There is a deeper limit than paperwork.
The stomach is the easy organ, because it can be filled with water and swum through; the small bowel and the colon offer no such pool.
And a camera that finds something cannot do anything about it, so a positive result still sends the patient to a real endoscopy, as an engineering profile of the project makes clear.
Why anyone is funding it anyway
Roughly 80 percent of upper endoscopies turn up nothing significant.
That is the actual target: not the sick patients, but the four in five who go under anesthesia so a physician can confirm an absence.
Strip out the hospital, the sedation and the recovery room, and the cost of looking inside a stomach drops toward the cost of a blood test, which changes who gets looked at rather than how they get treated.
The company’s stated pipeline runs from a surgical version to coordinated fleets, which is a long way from a camera that can barely radio home.
Every one of those steps is a version of the same problem that tiny fliers face in air and that a flippered swimmer faces in water: at small scale, the physics you grew up with stops applying.
The robot pill works, in the sense that it swims and it sends pictures, as trade coverage has tracked for years.
Whether it reaches a patient is not an engineering question anymore.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.
