Grind the blockage from the inside.

A 61-year-old woman with diabetes has felt full after a few bites for a month. She’s nauseated most mornings and has lost 9 pounds. Her gastroenterologist finds the reason on an upper endoscopy: a dense ball of undigested plant fiber, about the size of a tennis ball, sitting in her stomach.

The plan comes in stages. First she drinks cola for a few days, a surprisingly well-studied way to soften these masses. Then comes a second endoscopy, under sedation, to cut what’s left into pieces with wire snares. If the pieces are too tough, or the mass has moved into the small intestine, the next step is surgery.

Her stomach doesn’t empty well, so there’s a good chance it happens again.

A blockage with no good tool

Doctors call these masses bezoars. The most common kind, a phytobezoar, is built from plant fiber; persimmons are a notorious cause. Trichobezoars are made of swallowed hair. People who’ve had stomach surgery and people whose stomachs empty slowly, as many with diabetes do, are most at risk, and the masses tend to come back.

The treatments are all compromises. A 2013 systematic review in the European Journal of Gastroenterology & Hepatology found that cola alone fully dissolved about half of the stomach phytobezoars reported in the literature, and that cola plus endoscopy cleared about 91 percent. The persimmon kind is the hardest to dissolve. Cutting one up by endoscope can take several sessions, each one under sedation. Large hair bezoars, and bezoars that block the small intestine, usually end in an operation.

Bezoars are one of a family of problems in which something solid is stuck in the GI tract and needs to be broken down or moved along. Food that lodges in a narrowed esophagus is another. For all of them, the tools are the same: wait, dissolve, scope or cut.

What I’d build

I’ve designed a different tool: a capsule 9 millimeters wide and 22 millimeters long, about the size of a large vitamin, with a permanent magnet inside and a small spiral cutter on its nose. The patient swallows it with a sip of water. A larger magnet outside the body, held by a robot arm, does the rest.

The idea leans on something that’s already in clinical use. Magnetically steered capsule cameras have been cleared for examining the stomach; AnX Robotica’s NaviCam system received FDA clearance in 2020. Steering a swallowed capsule with an outside magnet works. What’s new here is asking the capsule to do mechanical work when it gets there.

1. 01 · SwallowNo sedationThe patient swallows the capsule. Ordinary swallowing carries it down in seconds.
2. 02 · SteerA magnet outsideThe outside magnet follows the capsule and keeps it pointed at the blockage, away from the wall.
3. 03 · GrindTorque, not forceTurning the outside magnet spins the capsule. The cutter grinds the mass into pieces small enough to pass.
4. 04 · WatchIt stops itselfThe console tracks the torque, tilt and spin rate it measures from outside, and halts the field the moment any of them looks wrong.

The physics drove the design. A magnet outside the body can barely push a capsule this size: in my simulations the push came to about a tenth of a newton, too little to shove a packed mass anywhere. But the same magnet can twist the capsule hard, at tens of millinewton-meters. So the capsule doesn’t try to ram the blockage. It grinds it.

What it did in simulation

The capsule exists today as a design and a physics simulation, and I want to be precise about that. The simulation runs in SOFA, an open-source framework for medical simulation developed at Inria in France, and models the GI wall as soft tissue that stretches, presses back and can be cut. The body is BodyParts3D, an anatomical model of an adult man built from real scans by Japan’s Database Center for Life Science.

I started with the hardest place to work: the lower esophagus, where the tube is narrow, the wall is about 3 millimeters thick and there’s no room to steer around anything. A piece of cooked beef sits lodged above a 13-millimeter ring just above the stomach. If the capsule can grind there without hurting the wall, the stomach, where bezoars form and where magnetic steering is already proven, gives it far more room.

From the October 2026 simulated run in the human model. The wall there is about 3 millimeters thick.

The swallow worked the way it should. The capsule reached the blockage in about 8 seconds, stayed within 10 degrees of the center line the whole way, and started cutting as soon as it touched the meat. The magnet followed it down the back, 18 centimeters from the capsule, and kept it spinning at 6 turns a second.

The safety side held up too. The console learned what the capsule’s torque looked like spinning freely on the way down, about 2 millinewton-meters, and watched for any rise. It never had to stop. Through the part of the run shown, the cutter touched the wall less than a twentieth of a millimeter deep.

The run also found the next thing to fix. The cutter ground a pocket into the front of the meat, about a seventh of it in the stretch shown, and then the capsule’s body, which is wider than the cutter, wedged against the rim of its own hole. The next version’s cutter is as wide as the capsule, so it can follow itself in. After that comes the stomach: a fibrous bezoar, a bigger space and a magnet steering from the front of the abdomen.

The objections

What stops it from cutting the wall? Three things, by design. The cutter sits recessed inside the capsule’s outline, so the body touches the wall before the blades do. The torque is small, a few millinewton-meters, and capped by the strength of the outside magnet; the capsule can’t twist harder than the field allows. And the console stops the field if the load rises, the capsule tilts toward the wall or it falls out of step with the magnet. Proving that margin on real tissue is the first job of the bench study below.

What if the mass won’t grind? Then the capsule has done no harm, and the patient goes on to endoscopy, which is where they were headed anyway. It doesn’t close off any option.

Isn’t the stomach too big to work in? That’s the problem magnetic capsule cameras already solved. They hold and turn a capsule in a stomach distended with water. Grinding adds a load to push against, and the bezoar itself is that load.

What happens to the capsule after? Its magnet lets it be held, moved or checked from outside until the work is done. Then it passes, as capsule cameras do.

How I’d find out if I’m wrong

I’d prove it in stages, and each one has a number that can end it.

• First, the bench. Build capsules and a robot-held magnet, and test them on explanted pig stomach and esophagus with bezoar models made from plant fiber, and with cooked meat lodged above a 13-millimeter ring. It passes if it breaks down at least 8 of 10 within an hour and tissue sections show no injury past the inner lining.
• Then an animal study. In anesthetized pigs, with a bezoar model placed by endoscope, run the capsule through the same protocol. It passes if it clears at least 8 of 10 with no perforation, and if an endoscope afterward finds nothing worse than surface scrapes.
• Then a first study in people, under an FDA investigational device exemption, in patients who would otherwise be scheduled for endoscopic fragmentation, with a gastroenterologist and a scope in the room for every case.

If any stage misses its number, we stop and say so. If it works, the payoff is a treatment a clinic could run without sedation or an operating room, for patients who today face repeat procedures and sometimes surgery.

For the woman with the tennis-ball bezoar, it would mean swallowing a capsule in a clinic chair while a magnet on an arm settled against her abdomen. If the mass came back a year later, she’d swallow another.