I've been involved with several small startups. My feeling is that in the first few weeks or months of a startup's life, your main goal is to make sure you're running in the right direction. You're about to release something as quickly as possible (minimum delightful product) to the market to see if you're on the right track.
That means that, you're mostly in chaos. you don't have time to learn new skills. You hire the best people and pay them well so you don't have to worry about teaching anything. It's about running as quickly as possible and wasting little money in the process.
That means that to me, the learning phase only happens after the startup has matured enough to be in the right direction. after the second funding phase, probably, it has time to worry about things like quality, and growing the people.
That's why I would not recommend TDD either when you're just trying to gauge people's reaction to something you might throw away in two weeks.
I think that's also the reason why in many established companies who started as startups, the code base is so crappy - because the initial idea was to get something out there. It's fixing the code base that considered part of the "learning phase". But unfortunately, many companies continue to exist in chaos even after initial startup phase, and don't give enough time for skills and practices to be learned by the team.
But that's a whole 'nother story.