Learn several very different programming languages. Make sure to include at least lisp, c, a scripting language (eg perl or php,) a parallel language (eg erlang or mozart-oz), something funky (prolog's nice this time of year,) and sql. These will give you a set of semi-compatible worldviews that'll let you start doing a better job of choosing tools than blub. If you dig in hard, you can do a pretty decent job of learning the wheel ruts of each of those in two months.
Spend the following year crapping out weekend projects to learn fast. Use that time to see what reaches customers, to see what breaks under load, et cetera.
After that you'll be in a much better position to take on founding.
I don't agree with this. He's not trying to become an all-star programmer, he just wants to learn how to build a prototype to get a startup off the ground. He should just jump right into what he will likely use. That's probably going to be Ruby on Rails or similar.
It's a little hard to believe that you'd be comfortable building your future on something you've only just learned to do, without getting to know the basics of your toolchest to the point that you can even make decisions well.
All I know is that when I've hired, the number of genuinely different languages that someone speaks has been a very strong predictor of whether their code can be kept in the long run.
But sure, picking a language without even knowing what the product will be? That could work too.
I know plenty of amazing devs who do extremely well either as an entrepreneur or as an employee who know none of what you originally suggested. I agree with Sean. He just needs something to start off with.
Spend the following year crapping out weekend projects to learn fast. Use that time to see what reaches customers, to see what breaks under load, et cetera.
After that you'll be in a much better position to take on founding.