For those of you wanting just to play single soundfiles or record a short snapshot (in a single buffer), the NDK project will be enough.
I had a look at the "Android NDK Beginner's Guide" and it really does not explain too much more than the sample NDK project.
But really, for these applications, you are probably better off with one of the Java APIs. Using OpenSL ES for them is a bit of an overkill IMHO. There will be much more code to write and no real gain.
The advantage of OpenSL is for those interested in lower latencies (when these become available) for audio synthesis and processing, and for this, you will probably want to stream audio, rather than play/record single buffers. The book chapter and the NDK example will not teach you about that. The blogposts in the audio programming blog are more informative for these purposes.