I think the answer here depends somewhat on what type of new user we are talking about.
For myself, when I want to see what a language can do I want to write some numerical code for a “standard problem” completely from scratch. This will indicate to me whether the language will actually be usable when I’m writing complex code for real applications. This approach is basically what led me to Julia in the first place: as you can imagine, I was always extremely uncomfortable with Python. If this is what trying out a language means to you, by all means I encourage you to download a nightly and dive headfirst into 0.7, and I also dare say, you are in for quite a treat with Julia. Then, when you need to use packages you can take a step back and write 0.7 compatible 0.6 code (if necessary) as @ScottPJones suggested.
Alternatively, a lot of people just don’t care about whether you can write things “from scratch” or even modify existing code and there’s nothing wrong with taking this approach, it’s just a different use case. Anyway, if that’s you, then I wholeheartedly agree with @ChrisRackauckas that you should start with 0.6 and wait a few months after the release of 0.7 for the ecosystem to catch up.