Nice! With this, you can get the same parallelizability and fast incremental updating as a hash tree, with the same output as a simple hash-of-all-the-bytes.
Yeah, exactly. I am using it for hashing btrees, will put this up next week or so on github.