I don't think my advice has gotten any stupider. If anything, probably better, because I've seen so many startups' paths now.
It's funny you should use the word stateless, though, because one of my tricks for scaling is to keep all the state on the client. E.g. I tell the startup "remember to tell me x next time we talk." I've been doing that for a while, and it seems to work fine.
Networking between the startups has actually increased. The more startups there are, the more likely you are to be in the same batch with another one that would be a natural beta user for your product, or that happens to be expert in some technology you have questions about. Maybe there would be problems with startups connecting if there were 1000 in every batch, but between 8 (summer 05) and 26 (summer 09) things just get better.
Earlier on this year I was having dinner with a bunch of start-up founders in London (all non-yc I believe, but well connected with the yc community), and your comment on statelessness was bought up, I think you managed to get it stuck in quite a few heads.
Are office hours a good measure of advice? What about the dinners?
Is the advice really completely stateless?
Some of the benefit of YC comes from networking between the teams. I think this benefit may diminish as the group grows.
Just thinking up objections.