Just read Paul's
new article. It's a good read, warmly recommended.
Most of what he says is the usual common sence. The most interesting idea is his noting that it doesn't take a brilliant patent to start a startup. You merely have to make a good product, better than all your competitors. The example of Google is given: there were quite a few web search companies when those came, and yet they won.
Eli Bendersky's website