One thing that worries me about "stealth" start-ups is that it implies (not all the time, but often) a lack of market research and validation ahead of time. While there are ways to get potential customers excited about a product and challenge your assumptions without necessarily revealing everything, more often than not, it seems like the majority of the assumptions will not be tested until launch, which ends up being a few hundred thousand (or million) too late. [0][1][2]
There are some products that don't necessarily need that level of validation ahead of time. There are some founders who have networks so wide, that they can start with a userbase of 5,000 people regardless of what they build. But for everybody else, proper validation of your assumptions might be the one thing saving your start-up from failure. And not telling anybody about it seems like a pretty easy way to get there.
[0] http://www.businessinsider.com/33-startups-that-died-reveal-...
[1] http://www.forbes.com/sites/ericwagner/2013/03/19/warning-7-...