What a startup can do is focus on a technique, rather than a specific discovery, and use that technique to make discoveries. The example I'm thinking of is Siluria, an MIT spinoff (or, at least, involving people from MIT; they were based in the bay area), which used phage display templating to make catalysts. They came up with the first practical catalyst for the oxidative coupling of methane to produce ethane. I believe they've since been bought by a larger petrochemical company.
That might be good. I guess one problem with that is that if the technique is not published or even sold, then others can't use it to make discoveries.
