Even simple things like trying to breed out deleterious recessive genes are incredibly hard.
The problem is that there are lots and lots of these recessives, but each recessive is fairly rare. Say there is some really bad genetic disease that affects one in 40,000 people. If we decide that these people should be prevented from breeding so they can’t pass on their bad genes, what have we done? Not much. The problem is that if the gene is expressed with frequency 0.000025 and you need two copies of the gene to express the trait, the trait is expressed in only 1/4 of the offspring of two carriers. that means the frequency of the gene in the general population is the square root of (0.000025 *4), or one in 100. That means that one person in 100 is a carrier for the gene, but only one person in 40,000 expresses the trait.
That means that in order to completely eliminate the trait you’d have to prevent 1% of the population from breeding. If you only eliminate those who express the trait, you only make a tiny impact on the frequency, since carriers are 4000 times more common than those who express the trait.
And that leads us to the next problem. There are hundreds of such genes. If you want to eliminate genetic disease you’d have to prevent carriers of all these diseases from breeding. But almost everyone has several deleterious recessives. You’d have to prevent almost everyone from breeding.
So if the goal is to eliminate genetic disease in the general population, it’s nearly impossible because you’d have to eliminate most of the general population. If your goal is merely to create a new breed of humans it’s doable, because you can select your founder population and cull any carriers.
But the next problem in creating your breed is the criteria you use to establish your breed. What exactly are you trying to accomplish? Create a race of supermen? OK, but you’d have to define what it means to be a superman. And all the selective breeding we’ve done on animals hasn’t created superanimals. There’s no breed of superdogs, unless you define “super” narrowly. There are breeds of dogs that can do things that no wolf could do, but none of these dogs are obviously superior to wolves. Greyhounds are faster than wolves, bloodhounds are superior smellers, dachshunds are better at hunting badgers, labradors are better swimmers, mastiffs are stronger, and so on, but no dogs are superwolves. Even breeds like labs that are very well rounded and very good dogs for all sorts of situations aren’t “super” unless you define “super” as “superior ability to get along with humans”.
And even if you establish a long term breeding program for humans, how do you keep this program going for multiple generations? What’s the benefit for the breeders? Livestock breeding programs continue across multiple generations of humans because new generations of farmers want improved livestock. But what is the goal of the human eugenics program? Once the masterminds of the program have died what’s the incentive for the next generation to continue the program? Of course, such a program is incompatible with a liberal democratic society, because the subjects of the breeding program wouldn’t be allowed to control their own breeding. You’d have to have an authoritarian government. And if the breeding program doesn’t meet the needs of successive generations of dictators, it’s going to be dropped.