The weak point in any encryption system has long been: how random are your random numbers? Providing special hardware for generating random numbers, special CPU instructions etc has always been open to the objection that their randomness could be subverted -- whether through some minor stupid mistake in implementation, or deliberately by certain powerful agencies -- and you would never know. Now, a new research paper <https://www.theregister.co.uk/2018/10/01/hardware_random_numbers/> proposes that normal CPU operation itself can be a source of random numbers. And the way these are obtained would in fact be very hard to subvert, since it is based on detecting subtle timing variations in the operations of various circuits. In other words, it is obtaining randomness from hardware that is not designed to be working randomly at all. And this should -- hopefully -- make the randomness very difficult to compromise.