I've done this (and successfully exited a startup) and I think you should wait. A bit.<p>Startups generally take a lot of capital, and you do want to make sure you are building up to some level of savings. If you can do something as a side project and have it grow to start making money, that is super great - though be prepared that there is some risk in terms of seeking outside investment (and that this also takes a large amount of effort).<p>As such, work in a couple of startups as well as at least a medium size company (several hundred people or so) and get a feel for how they do business. Learn what works and what doesn't.<p>Finally if you do decide to go through with it, be aware your life and company are going to pretty much merge, which can be more mentally challenging than I can prepare anyone for - which will affect you even after you're done with everything.<p>I think the most important skill you need to learn is instinct, which comes from seeing a lot of things happen in both good ways and bad ways.<p>Keep your idealism - build amazing side projects, hone your craft and have fun. There is no timer that says you have to start something only in your 20s and software development in the wild is a lot more different than college.<p>Keep those ideas going, but don't feel rushed if you don't want to do something right away.<p>Try to find some cool startups to work for, and learn from them. That may be enough for you creatively. If you still think you have some revolutionary idea, go for it. But don't approach it as "I want to do a startup", you should have some idea that you really really like and want to do for a very long amount of time, and have a lot of ideas around.<p>A lot of the attempts to start startups fail because the idea is principally around "I think this could make money" because we glamorize startups, when we should look at many of them as just risky ventures between a handful of people.<p>That's ok, but you need to eat. And you need to have hobbies and live your life too.<p>There's plenty of time.