Hey thanks for the great questions! Happy to elaborate on anything below if it's helpful.
1/ It might be easiest to define in terms of suitability for a particular application. As you probably already know there are a number of industry tests which are useful in evaluating the effectiveness of a random number generator (Diehard, NIST, etc). Our service is built on top of industrial-grade hardware that passes these tests and is suitable for use in cryptography.
2/ Great question. It's easier to reason about in the context of specific use cases. For certain applications it increases trust to outsource RNG to a neutral third party that doesn't have a stake in the outcome.
I'll also add that we're considering open-sourcing our implementation of raw hardware bytes => data as a way to build trust and transparency in how we're generating data.
I always think of this blog post when these posts come up about generating random numbers.
http://gamesbyemail.com/News/DiceOMatic
There is a link to ver 1 built out of Legos.
1. What do you mean by "true random"? Aren't the bits generated by a hardware pseudo random at best?
2. Why and how should users trust the randomness of a third party solution that they do not control or see the implementation of?