* First off, doing a startup is not just about sitting in a dark room and writing code without a boss. It isn't even mostly about that. Your average engineer has no idea how much non-technical crap there is to do, and consequently has no idea that the things they like to do don't strongly overlap with the set of things they think they'll be doing as a startup founder. It isn't just pizza and coding sprints -- it's about talking to customers and coordinating with other people and marketing and finding photos and making spreadsheets and raising money and bookkeeping and filing idiotic paperwork with bureaucracies and diffusing personality conflicts and a million other things that good developers never think about and don't ever want to do. And if you succeed, your reward is more of that stuff and less coding.
* Second is that you'll work like crazy and have to bear a lot of stress. That's the way it goes. Most people aren't cut out for that lifestyle. Doe-eyed college graduates, especially, haven't ever done a project that lasts more than a semester or two. They have no idea what they're getting into.
* Third, if you don't have a co-founder, and you don't have any experience, you're doomed. Go work for someone else for a while. I know, I know...$FAMOUS_PERSON did it. They're the exception, not the rule. If you're in that situation, there's a litmus test: is your product currently built and exploding in popularity? Then you may be the next Zuckerberg. Otherwise, go work for someone else for a while.
* Fourth, these days, most people confuse "doing a startup" with "making a cool feature". They're not even close to the same thing. I can't tell you the number of resumes I've seen from developers who made a half-baked marketplace website that they called a "startup", then gave up when the users didn't come and they realized that 99.999% of the work was ahead of them. Building the product is the first step, not the last one.
* Fifth, just because it's cheap to rent server space and open-source means that you don't have to write CGI by hand doesn't mean the competitive landscape is easier. In fact, it's harder. For any idea you come up with, there are either 50 sophisticated, me-too competitors, or there will be within 10 minutes of you getting any publicity. The flip side of everything being cheap is that any bozo with a credit card can do it.
...I could go on and on with this. Point is, most developers have this hazy, romantic, idyllic notion of "doing a startup" that bears no resemblance to the actual thing. Communicate this honestly and clearly to someone when they're waffling between taking a job and going off on their own, and you've got chances. After all, anyone who's really ready to do a startup isn't going to be talking to you about a job. They're going to be doing a startup.