Expanding on my comment here, I think the most illuminating way of looking at the debate about how big a society (ha!) needs to be to support certain levels of technology may be to look at some natural experiments. Specifically, we know about a number of cases where societies have decided to acquire complex new technologies with limited outside help. Basically, these are clandestine weapons projects.
Now, most if not all of them had some degree of outside help. But the question is really how much you can do with the equivalent of taking along a library on the space ship. To some extent, getting outside help is analogous to this.
Pakistan, for example, succeeded in developing a clandestine nuclear weapons program. Much of the information it needed was essentially canned - they could study it up. The gating factor was, as always, the fissile material. Having tried the relatively easy plutonium route and been caught, they proceeded with highly-enriched uranium. This meant that the technology barrier was designing a working centrifuge and then building enough of them to scale up. A lot of people over on the Crooked Timber thread think, essentially, that this is the difficult bit - there's a lot of implicit knowledge embodied in the process that you can't get from textbooks.
An example of this is the performance of the Iranian enrichment cascades. There have been repeated instances of them seeming to progress much more slowly than the known capabilities of the R-2 machines, and over at Armscontrolwonk, you can argue endlessly whether this represents a policy decision to go slow or else operational problems due to their inexperience.
However, arguably, Pakistan did use a textbook - A. Q. Khan brought over information from URENCO that helped enormously. The rest was a question of learning by doing, or kaizen - continuous improvement. Interestingly, Khan's private nuclear trading operation essentially sold the same sort of thing, a sort of starter-kit of centrifuge parts and documentation that let his customers start to learn about enrichment operations.
The biggest counter-example is North Korea, which did get a lot of outside help in the 90s for its missile program. Rather than just getting documents and example devices, North Korea imported whole sections of a rocket engine production line and many of the people who ran it. They may not have stuck around long, but it remains true that the North Korean nuclear and missile development projects started off with what could be described as on-line outside help. They didn't just have the documentation - they could ask the experts. But their achievements are significantly less impressive than Pakistan's.
Another case is the development of long-range drug smuggling craft. Recently, the Colombians found the first known drug sub capable of submerging fully and also of making a trans-Atlantic voyage. It is, of course, a mystery whether any others are operating. The interesting bit is that it seems unlikely that their builders have access to North Korean-style on-line help. It's just possible they managed to find and recruit a submarine designer, I suppose. But there's no evidence of that. What there is evidence of is kaizen; for years, they have been building progressively more impressive and capable craft, from boats with a low freeboard, to semi-submersibles, to bigger and longer-ranged semi-subs, and now to a full ocean-going submarine. That would suggest that they have general shipwright's skills and heavy metalworking, and they've progressively learned more as they went.
What conclusions? First of all, don't underestimate the power of general purpose technology. (This is essentially the promise of Ted Nelson's Computer Lib: The computer is the most general machine we have ever made. You can and must understand computers NOW...) Second, don't be obsessed by outside help/state sponsors/whatever. They're a way of denying other people agency.