When a car hit San Franciscan cyclist Seth McGinnis, he flew seven feet into the air before landing on his wrist. He broke it in two places, chipping his elbow in the process.

After weeks of wearing a traditional cast, McGinnis was miserable. It smelled. It itched. He couldn’t shower. He couldn’t swim. He couldn’t exercise properly. He wanted something different—something light, breathable, comfortable.

That’s when a friend told him about a Silicon Valley firm, Autodesk, that wanted to design a new kind of cast. But McGinnis’ doctors were skeptical.

“It was at that point I started advocating for my own health care,” McGinnis said. He went against his doctor’s advice. He had a friend cut the cast from his arm using a handheld rotary saw, and opted for something completely untested: a cast created through a collaborative effort between man and machine.

Bradley Bermont

Cue Arthur Harsuvanakit, a designer for Autodesk, whose team turned to the company’s generative design software for the project. They call it computer-aided design, but it’s more like an artificially intelligent engineer that collaborates with a human counterpart. Harsuvanakit described the process like this: “We scan the person’s hand,” using a 3-D scanner, and “we identify where exactly it needs the most support.”

He then designed a traditional-looking cast and input that into the software, noting the areas that needed the most support. In McGinnis’ case, that meant both sides of his left forearm and thumb.

Harsuvanakit’s design had to be optimized, removing material to make it lighter and more efficient without sacrificing support or structural integrity. “It’s a bit of a black box,” Harsuvanakit said about the optimization process. “More times than not, it breaks your assumptions on the solutions you think your setup will provide.”

The software runs off an internal feedback loop, using sophisticated physics models to figure out the best way to achieve the goals set by the human designer. After testing hundreds of designs in less than an hour, the software presents a handful of the best. Harsuvanakit plays curator and editor.

“It’s kind of a back and forth between the computer and the designer,” he said. “You’re editing and changing design direction based on what it can give you.”

The goal is to make the strongest, lightest, most functional design possible. When Harsuvanakit sees a design he likes, he pulls up the model and tells the computer to further optimize the shape and solve for a new variable.

For McGinnis, they started with the most supportive cast possible. Then, Harsuvanakit and his team solved for other variables like weight, breathability, and comfort—each iteration subjected to McGinnis’ feedback. Early in the experimental process, each version took about a week to create. By the end, however, it was less than a day.

With every result, Harsuvanakit would make a few changes to the computer’s design and feed it back into the software, where the process would repeat itself. When Harsuvanakit was satisfied, he would 3-D print the finished product.

The result was a ready-to-wear, space-age looking, hard plastic cast with branching supports weaving along the top of McGinnis’ forearm like tree roots, connecting two natural bridges on each side. Harsuvanakit and his team included two hinges to make the cast removable, which was the only way McGinnis could use the cast since it couldn’t be printed directly onto his body.

In total, the team went through seven prototypes before settling on a final design.

That's not to say the process was completely smooth. Harsuvanakit originally designed the prototypes to fit McGinnis’ right arm, even though the left one was the one that was broken (they didn’t want to take him out of the cast before they had a functioning prototype). But the final product—they simply mirrored the cast for his broken arm—didn’t fit. McGinnis had to resort to electrical tape to put the traditional cast back together and wear it for one more day.

“The first day I went swimming in the pool and did laps, I was blown away at how nice it felt,” McGinnis said. Between waterproofing, breathability, and accessibility to any itchy areas—no coat hangers required!—he only wishes he could’ve gotten it sooner.

McGinnis only wore the cast for the last month of his three-month healing process. Since his doctors declined to participate, we can’t know for sure that it’s medically viable—though McGinnis’ physical therapists say that he’s recovered well and McGinnis himself says that he healed faster because of it.

From Autodesk’s perspective, the experiment was a rousing success, adding to their portfolio of generatively designed products that already included furniture, power tools, and engine blocks.

Still, Harsuvanakit admits, the technology isn’t ready for widespread medical use. “It’s more of a proof-of-concept at this point,” he said. He noted that the iteration time would need to be quicker and the software needs to be more streamlined. Next steps include getting the backing of a healthcare provider for financial backing and continued experimentation for FDA trials.

Right now, “it’s a great talking piece for the health industry,” Harsuvanakit said. But he knows it will be more.