There was a certain kind of quiet that settles over a house during a lockdown. For most families, it meant sourdough starters and Netflix queues. For one Virginia teenager, it meant clearing off the basement ping-pong table and turning it into a robotics lab. Benjamin Choi was in tenth grade when the world shut down in 2020. His summer research placement, studying aluminum-based fuels, had evaporated overnight. What he had instead was time, and a memory that had been sitting with him since third grade.That memory was a television segment about a paralyzed patient who could move a robotic arm using only her thoughts, thanks to sensors surgically implanted in her brain. It had stunned him as a kid. It had also bothered him. The technology worked, sure, but it demanded a dangerous operation and a price tag that put it out of reach for almost everyone who might need it.So, with a 3D printer that had cost his sister $75 and a spool of ordinary fishing line, Choi set out to build something better: a prosthetic arm that read brainwaves without ever touching the brain itself.How she used small printer to solve big PproblemThe printer he had access to could only produce pieces a little under five inches long, nowhere near enough for a full arm. His workaround was almost comically low-tech: print the limb in dozens of small sections, then bolt and rubber-band them into a working whole. The first prototype took roughly 30 hours to print. It moved using head gestures and basic brainwave signals, and once it worked, Choi published the build instructions online so anyone else could try it too.He wasn't exactly starting from zero. Years on competitive robotics teams, a hobby that had taken him all the way to world championship stages, had already given him a feel for building things that move. Coding was newer territory; he'd picked up Python and C++ largely by watching programming tutorials on his own time in ninth grade.She worked on 75 versionsThat first prototype was just the beginning. Choi kept refining the design, more than 75 versions later, the arm had evolved into something built from genuine engineering-grade materials, strong enough to bear loads of roughly four tons. And the manufacturing cost had barely moved: around $300 (nearly Rs 3,000).For context, that's a fraction of what's currently available. A basic, body-powered prosthetic arm typically runs about $7,000. The most advanced option on the market, a robotic limb with 26 joints and hundreds of embedded sensors, paired with nerve-rerouting surgery so patients can feel texture through it, has carried a price near $500,000.Choi's version skips implants entirely, relying instead on electroencephalography, the same non-invasive technique doctors use to diagnose conditions like epilepsy. Two electrodes do the work: one clipped to the earlobe as a baseline reference, another on the forehead reading electrical activity from the brain. That data streams via Bluetooth to a microchip built into the arm, where an AI model Choi designed himself interprets the signal and translates it into movement. Head gestures steer the arm; a deliberate blink brings it to a stop.Teaching a Machine to Read a MindTraining that AI meant real, hands-on data collection, no easy feat for a high schooler working alone. Choi recruited six adult volunteers and spent about two hours with each of them, recording brainwave patterns while they repeatedly clenched and unclenched a hand. Feed enough of that data into a model, and it starts learning to tell the difference between "move" and "don't move" just from the electrical noise of a thinking brain.The resulting algorithm is substantial: more than 23,000 lines of code, underpinned by nearly a thousand pages of supporting mathematics and seven entirely original sub-algorithms. In testing, it identified user intent with roughly 95 percent accuracy, well ahead of the 73.8 percent that had previously stood as a benchmark for comparable neural networks.The Recognition Followed the WorkBy 2022, the project had landed Choi among the top 40 finalists at the Regeneron Science Talent Search, one of the most competitive science fairs in the country for graduating seniors. Funding and mentorship arrived from MIT the year before, giving him room to spend roughly six months exploring cloud computing so the arm could eventually connect to the internet. A Simons Fellowship brought him to Stony Brook University, where he collaborated remotely with an electrical and computer engineering professor on refining the underlying machine learning. Awards from the Regeneron International Science and Engineering Fair, the Microsoft Imagine Cup, and a national at-home STEM competition followed, along with a manufacturing grant from a 3D-printing materials company in late 2020.Still BuildingThe story hasn't stopped at the science fair stage. Choi went on to do machine learning research at Johns Hopkins' Applied Physics Laboratory in 2023, spent the back half of 2024 working as a researcher with NASA, and now works as an AI researcher at Harvard's Kempner Institute. He's finishing a degree in applied mathematics at Harvard, alongside a master's in computer science earned concurrently, a pace that suggests the kid who once bolted printer scraps together on a ping-pong table never really slowed down.The arm that started it all keeps evolving too, its core signal-processing work refined further during his time at Johns Hopkins. But for Choi, the arm was never really the endpoint. It was the proof of concept that convinced him machine learning could solve problems that actually change people's lives, and that's the thread he's been pulling ever since.