An ordinary small business doesn't need people so smart that they can redefine the problem. They just need people who are cheap and reliable. Whereas startups generally try to hire the smartest people they can, even if they're difficult or expensive.
Software as a Service, and plain old ISV application software can both scale immensely. I do not need to hire extra bodies if I'm selling 1 million copies a day of my app than if I'm selling 10 copies. The only thing stopping an ISV from growing is the popularity of the software, just as the only thing stopping a Service from growing is popularity of their webapp. In neither case there is no roadblock to growth caused by a lack of resources.
ISV != small. It never has, and it probably never will.
Give me a break. If Intuit sells 10mm copies of QuickBooks a year, why don't they just have 5 or 6 employees instead of the hundreds that they have (that work on the QB side)?
Your "Build a better mousetrap and the users will come" approach just apply to every scenario. Maybe to yours, but not to everyone's.
Why do Google and Yahoo have 5000+ employees, when Craigslist gets by with about 20?
Many companies hire people because they can, and because cash in the bank gets a miserably low return. The extra employees aren't strictly necessary, though they do help (presumably over the marginal return of cash, which is about 0 when you figure in inflation).
BTW, Intuit's sales took off when they did only have 5 or 6 people. They probably wouldn't have been able to release anything else if they hadn't hired more people - the engineers (all 2 of them ;-)) were answering phone calls all day instead of coding. But they could've grown quite large just on Quicken.
There're a bunch of other ISVs that grew very large on very small teams, eg. Aladdin (StuffIt) and NullSoft (Winamp). Weren't many of Borland's early products (Quattro Pro, Turbo Pascal) also done by teams of about 4?
2) Your "Build a better mousetrap and the users will come" approach just apply to every scenario. Maybe to yours, but not to everyone's. No idea what you are going on about here. I never said anything about building better mousetraps, neither literally nor metaphorically.
3) I wrote a big long reply explaining why you're wrong, but Nostrademons said it far more clearly, so I'll leave his words as the definitive answer...
If you actually wanted your comment to be taken as a seperate thing, and not read in conjunction with your other writing, then I guess we don't have anything to talk about - as Eric defines small ISVs, they are effectively... small :-)
One of the reasons PG suggests building server-based software is that you have more flexibility in language choice, and hence more ability to hire great hackers. But there are other reasons to go for that business model: recurring revenue, fewer distribution hurdles, less chance of incompatibilities, ownership of customer data, network effects, etc.
I think there was an article about open-source software and Sun's cash hemorrhage that made it to Hacker News recently - it was either here or Proggit. That's another instance. Google was cited in the article as a company that supports open-source - well, that's because Google's business model is selling services supported by software, and so it behooves them to get the price of software down as low as possible.
For that matter - hardware, software, and software services are all complements (in a microeconomic sense), and so each of them has an incentive to commoditize the others. Hence we have Google supporting open-source and buying commodity hardware; Sun and RedHat supporting open-source; Microsoft training lots of developers and providing plenty of tools for them; Apple basing OS X on open UNIX instead of proprietary MacOS 9; and great hackers everywhere supporting open source, because it drives up the value of the service they provide.