Thanks for your fast response @knutwalker. I read through your reasoning with the code and believe you are correct.

The code looks to me like the caching doesn't work for a _1() will that returns null, i.e. repeated calls to memo() will repeatedly evaluate _1(), contributing to the quadratic execution time that memos were created to avoid.

Maybe the condition should only be based on v == null -- something like this?

public A _1() {
  if (v != null) return v.get();
  synchronized (latch) {
    if (v != null) return v.get(); //check again here in case v was assigned asynchronously
    A a = self._1();
    v = new SoftReference<A>(a);
    return a;
  }
}

I created #108 to deal with this.