I don't think that's right. If 50% of people became replaceable by robots, the other half doesn't become more valuable. Rather the value shifts to the robot owners.
The article paints a simplistic picture, which is probably not correct even to first order, but it may be a useful viewpoint. The stratification should only increase, with the super devs obsoleting the ordinary code monkeys, but the time frame on that might be large.
The value shifts to the robot developers. They are the ones that capture more of the value originally pooled by the entire workforce.
The robot owners get value in that they can lower their prices because of the replacement of labour with robots, and overall, society benefits because whatever the robots are making is now less costly (ie, takes less of your own hours working to purchase).
The key to understanding all of this is productivity. Most people tend to think in terms of raw job numbers, rather than productivity numbers.
Employing robots results in a more productive economy because the same (or greater) amount of output is achieved with less human input. This does release labour from that activity, but it also creates opportunities in other areas, as the savings people make on purchasing the robot-cheaper goods can now spend in other areas, which should create jobs in other areas.
Of course rapidly advancing technology produces lumps of unemployed which take time to retrain and redeploy themselves elsewhere, and it sucks to be that person. But eventually they will find employment, assuming something else isn't working against that (like oversized debt accumulation, which works against new enterprise creation by causing a negative investment flow, but that's another story)
You hypothesize an either-or where it's a both-and. The robot owners need somebody to program the robots, too, or they're expensive-yet-worthless hunks of metal. And if that talent is scarce compared to demand, the price is going to go up.
Maybe I'm nitpicking, but I'd say that effect is due to an increase in demand for robot programmers, not due to a decrease in the supply of the productive workers.
The article paints a simplistic picture, which is probably not correct even to first order, but it may be a useful viewpoint. The stratification should only increase, with the super devs obsoleting the ordinary code monkeys, but the time frame on that might be large.