In the contexts I usually see, the complaints of magic are regarding Ruby's fondness of Spooky Action at a Distance. I would then argue that Ruby or its frameworks do seem to rely too much on this "magic" that otherwise is condemned by recent literature (and Code Complete).
Now Rails on the other hand relies on a good deal more magic than say Django, but it's recognized in the community as a tradeoff and it is done carefully. Some of the things that rubyists accept in Rails they would never put in an application or library because it is deemed okay only by massively adopted convention at the framework level, but not generally a good practice.
Then you have libraries like rspec which started out being full of magic but which gradually got refined and cleaned up so that in rspec 3.0 you have zero monkey patches; it's just pure ruby, providing a totally modular set of DSLs that work together nicely and allow for amazingly readable test output.
That's known as Uniform Access Principle[0] and is a wonderful feature to have.
Ruby is far less magical today than it was 5-10 years ago. Magic is a clever word for "I don't understand this thing yet."
0 - http://martinfowler.com/bliki/UniformAccessPrinciple.html