If someone wants to claim "very small companies are more likely to try something
legally dubious because they have very little to lose," that's a completely different argument. Not to mention, YouTube and then Google had to pay up in order to square things away and neutralize the risk. Big companies are more comfortable playing fast and loose with the rights of individuals, while individuals and very small companies are more comfortable antagonizing powerful corporations.
Start-ups are NOT the best way to develop software. Start-ups are inherently inefficient. On top of doing the development, you have to take care of all the paperwork and technicalities of running a business. YC makes you move, which is a big interruption (you should only be moving if you NEED that interruption to leave behind all the distraction; some people are already coding like crazy and having to rearrange their lives like that just gets them off-track). You are trying to market yourself. The list goes on and on; I can't find the thread but somewhere PG said he only spent 20% of the time actually coding (my memory may be faulty, it might be 50%).
By contrast, working at Google you don't even have to cook -- they try to take care of all the minutiae. It's really no surprise that whatever they buy, the vast majority of the work is done post-acquisition.
The real purpose a start-up serves is as an advertisement that these people are very dedicated on that project -- when they are bought they are paid EXTRA to work on whatever they already wanted to work on anyway. [1]
THAT is the key difference. Most employees are hired to do some job they don't care about for some crummy wage, so no surprise productivity isn't high. If Google bought up start-ups and then reassigned everyone to other projects, nothing would get done. You don't get dedicated "in general", you are dedicated ONLY to what you're very interested in.
Hiring isn't obsolete; hiring people based on generality to do some unspecified thing for average pay is obsolete. It's simply a quirk that there are currently two models -- the useless traditional one and the "find people at another company who are already doing something and then pay them MORE to keep doing what they want to do ANYWAY" one.
It has always been obvious that letting people choose what they want to do and paying them more money for it yields better results than telling people to do something they probably have little interest in and paying them less money to do it.
Start-ups are a bad model for actual development because of the overhead. They are a good model for picking a certain type of people. I think PG just has this reversed.
[1] As a caveat, not everyone is dedicated to their technical ideas; they are dedicated to making a lot of money. Once they make the money, they stop working on the idea or anything like it. This seems to be what happened with ViaWeb, and would explain a couple of things: One, why PG is so insistent on "flexibility" and steering a lot of people away from their initial idea (because that's what he did -- except he wasn't passionate about the art gallery thing in the first place, so it's a lot easier to switch gears). Two, having steered applicants away from their initial idea, if they were dedicated to that, you've now turned them into people who are dedicated to the new thing for money only, and this could explain why YC's picks haven't been more successful. Your spirits also flag more when you're working on something you don't really believe in, which is why having a co-founder to lean on is more important. Plus there's the overhead -- how many of these start-up groups have a "business guy" as one of the co-founders? It's an inefficient ratio; imagine if every time Google hired a programmer they also had to hire a "business guy" for him.
I think it's pretty clear YC is much more compelling to PG than Yahoo! Store ever was. He could make a lot more money at it, too, if he doesn't steer the dedicated people away from what they are dedicated to.