I think the Indian AI ecosystem is growing rapidly. A lot of Indian entrepreneurs reach out to me seeking feedback about startups and products. And some of them have very interesting business ideas.
The problems that you see startups tackling are dramatically different in different cities. Silicon Valley is unlikely to produce the same set of companies as New York or Cleveland because the region has a different set of strengths and defining institutions.
The challenge in a startup is you hit a lot of turbulence, and you want people who understand that it's just turbulence and not a crisis.
Startup stories are always smoother in the telling than they are in reality. A startup is not one, but a series of 'Aha!' moments, and some which seem like 'Aha!' moments but turn out not to be.
You can fake a lot in a startup these days, what with Amazon Web Services and all sorts of off-the-shelf back-end components that let any even minimally competent duffer set up a Web app that does something. Intelligent planning for growth is rare among early startups, but it's the name of the game at a large, rapidly scaling tech company.
I think Yandex is something in between two different cultures. One originated from the old Soviet culture of the scientific institute. It was a free atmosphere of scientists, maybe too free because nobody cared about making money. Another origin is something close to what you usually see in California startups.
It almost goes without saying that when you are a startup, one of the first things you do is you start setting aside money to defend yourself from patent lawsuits, because any successful company, even moderately successful, is going to get hit by a patent lawsuit from someone who's just trying to look for a payout.