Agreed. This design is going to be very very hard to machine precisely and very expensive to prototype

They don't even say exactly what they did or how long it took or how well the tolerances turned out. They only give the classic understated view:

"Many parts of the prototype were built from high-performance plastics with high machinability and stable properties because the components of ABENICS have complex shapes, which makes estimating problems during prototyping difficult. Construction from strong materials such as metals will also improve the capability of the device."

Eh, they make precision stuff like this all the time. If they wanted to make a bunch they would first standardize the sizing then create production tooling for the grinding setups. Those gears would come out basically perfect every time.

The bigger problem is the output link is supported by the gear meshes. This means whatever load you put on it is directly supported by the small mesh contact patches. A more traditional system can have roller or ball bearing or bushing support.

Exactly my thought. They're hand waving all complex manufacturing to 3D printers. This will be a hard plastic gear for the foreseeable future.

Even more previous: https://news.ycombinator.com/item?id=28556208

Thankfully the open access model is becoming more popular, though I still find it insane that a lot of publishers demand a huge fee to publish articles as OA. I know they have to pay the bills and I guess it's better than ads, but they hardly even do anything anymore with online publishing and volunteer editors/reviewers.