Interesting, my perception is that it's gone the other way; that not only is YC now primarily accepting companies at a later stage when they've already been substantially derisked, but it's also investing in less risky companies to begin with. Several YC partners have said this as well, e.g. Jared Friedman:
"Companies are joining YC at a much later stage. When I started YC, most companies wrote their first line of code in the first week in the program. Today, many of the companies have been working on their business for a long time and some even have substantial customers and revenue before applying. If the companies in my batch applied to YC today, I doubt that many of them would get in." http://blog.jaredfriedman.com/2015/08/18/nine-years-of-demo-...
While I'd like to believe that YC is still willing to fund companies like JustinTV or Loopt, where there is substantial risk in terms of both the technology and the sociology, the reality is that if you go down the most recent batches of demo day companies most of them look like Dollar-Shave-for-X or Uber-for-Y. Maybe that's a slight exaggeration, but perhaps the most striking thing about the most recent batches is how few of the companies are obviously terrible ideas. Unlike back in 2006 and 2007, where probably a third of the companies failed to even launch, and another third probably never got more than 100 users or whatever.
Obviously YC is now also funding energy and biotech companies that are at least as risky as anything from 2005, but if you look at the verticals where they've been active the longest it certainly seems like the trend in those areas is toward later stage and lower risk.
I am sure this is a reflection of the companies they get to look at. If we use a dating analogy they seem to be getting more interest from supermodels, but this doesn’t mean they are getting more interest from potential life partners.
I think it's the opposite actually; there are an effectively infinite amount of startups, but the bandwidth of YC is (currently) finite. What we're seeing now is exactly what you'd predict just from the math, even without any data.
This blog post could be interpreted as a way of providing more information to founders, so that they can't then say "But I didn't know that accepting another accelerator would lower my chance of getting into YC." Now you do.
Also, it's a true statement that AirBnB and Dropbox succeeded because they were great companies, not because they went through YC. At least in the early days, YC had a very strong bias towards selecting companies that would succeed without them. You'd have to ask someone affiliated with YC if that is still true, but given comments from YC partners here, I suspect it is.
That statement is not mutually exclusive with YC being able to help companies that would've succeeded without them, and indeed, both Brian Chesky and Drew Houston have said that some of YC's advice was very helpful to them in critical early stages.