I'll extend Ted's answer one level, but I will also not give any references because I didn't research whether such argument is actually put forth by libertarians. (At least, in politics. I've seen it debated in software development). Nevertheless, it's a valid argument, to an extent, and I find it more compelling than the simple feedback mechanism known as "the invisible hand of the market".
In a word, development (especially long-term!) happens like evolution. In evolution, the feedback to the next generation is the traits and properties that "happen" to be aligned with successful social interactions (e.g. business practices). The good thing about evolution is that the "actors" (people, businesses, etc.) don't even need to know what caused the success (or failure). They just need to "inherit" such practices and, basically, do the same if they succeed and not do it if they fail.
Evolution is seemingly a very short-sighted process. It can't "plan" anything. Yet, as any research into evolutionary optimisation shows, it's an amazingly robust process, and on complex landscapes evolutionary methods nearly always beat deterministic optimisation. (And what can be more complex than human societies?) Evolution is the only known "natural" process that can create new information.
Now, for evolutionary model to work, several conditions must be true. First, there must be competition. ("Selection pressure" in evolutionary terms). The more actors, the better. Whoever loses must fail.1 Next, actors must be able to freely interact with each other and "borrow" successful traits. And that's it.
Generally, evolutionary process doesn't "like" rational interaction, like the elitist selection mentioned in the footnote1. It is quite counter-intuitive, especially for an "engineering" mind. It doesn't favour parsimonious, that is, the simplest and most "elegant" solutions. But this, it seems, quite aligns with libertarian mind.
1 "Must" should not be taken too literally here. Evolution is brutal but messy thing. It can be shown that in most cases non-deterministic selection (where successful individuals/practices simply have more chances to pass onto the next generation) is more efficient and robust than the so called "elitist" selection where exactly the best ones are always selected.