This article makes it sound like the main bottleneck in the first-run recycling process is sorting.

Could anyone with knowledge around ML, robotics and the like chime in on napkin math around novel labor-reducing approaches to improving the operating economics here? From the sound of it, this is not really a solvable problem short of some leapfrog technological advance.

Even then, you don’t want to spend more energy on sorting automation than you get back from the recycling. And this would only make sense if the diminishing returns after the first or second recycle made it reasonable to be issuing so many consumer plastics in the first place.