As a side note this is reason I've always believed building an business/app of any importance on-top of another site is just crazy. You're at their mercy and there's nothing you can ever do to stop them shutting you out.
This is just the 21st century equivalent.
Likewise, if you're building a destination website you're dependent on Google, email spam lists (don't get blacklisted!), browser vendors (MSFT and ESPN once conspired to make espn.com for IE-only), and even ISPs (see: Net Neutrality).
The risks are higher on Facebook and Twitter, but the upside is also higher. For example, I launched a Facebook application late last week and it already has 500k users. That kind of distribution is hard to get elsewhere.
Building on top of a business is sensible when you are depending on something that's been there for a while, know it's dependable and have a reputation. This twitter does not have.
If the same service as Twitter was provided by Google, Amazon or even Microsoft, you know it's backed by something solid.
That's why he's so upset that Twitter cut him off.
So, yes, building on Twitter is riskier. Building on Facebook is riskier. But as distribution channels they have huge upside.
If the risk is too high don't build and wait for the channels to become more mature.
Google is a multi faceted business, they have a good revenue model, have money, and are a proven company. Depending on them isn't crazy.
What's crazy is depending on Twitter, it's new, still have no solid revenue model to make money and sustain, in simple it's a single egg in the basket.
Don't trust the cloud. Appreciate its gifts but don't expect it will always be so.
And almost any business will be dependant, to some extent, on others. It's a question of how many of your dependencies are irreplaceable, and how likely they are to shut you out.
See for example http://en.gentoo-wiki.com/outage-10-08.html - they went down and lost all their data because the hosting company had a fight with the people who owned the building where the servers were kept. (Including backups, which made the host irreplaceable - with offsite backups, that wouldn't have been the case.)
Building on a site like twitter might increase the risk of being cut off on a whim, but it probably decreases the risk of competitors.