I believe the only reason you think that previous generations managed it is because previous generations of programming were tremendously less open in terms of visibility. I would even go as far as to say that if you could collect and objectively grade all code in the world 10 years ago and today, the collective code today might actually be of higher quality, simply because code is shared much more today. This leads to more knowledge, which leads to higher quality. In the past, that code was simply left behind on someone's hard drive.

Code itself doesn't grow into craftmanship, because craftmanship is what a person does. A person gets better with experience (both from doing and from observing) and inspiration. Greater visibility and sharing of code means better experience (through observation) and better inspiration (from people sharing their great ideas).

If you think I'm wrong, the alternative solution isn't difficult: you can avoid shared code repositories altogether and instead write more code from scratch. If you still want code sharing, you can start your own online repository where only thoroughly vetted code is allowed in. The important value to me is that people are not encouraged to censor themselves, but that we push ourselves harder to increase quality, by sharing both our knowledge and our work even more. If navigation is an issue, we need to build better tools.

Multiple choices, competition, and organic growth is healthy for the system – enforcing artificial limits believing it'll increase quality is not.