 Yeah, I like some useless things so the other day I was I was seeing that this work of two Google brain researchers they proposed this adversarial neuro cryptography in which basically they have two three neural networks Alice Bob and Eve that was in this kind of adversarial and setup and Alice and Bob had to communicate again with each other and Eve tried to break the system so I was a bit disappointed with the paper because in the end they didn't actually show the encryption algorithm that Alice and Bob had found so I was doing nothing so maybe I could try to do this so what I did I developed a simple neural network that I knew that could learn the one-time path and implemented this to see what could happen and actually the neural networks did not learn the one-time path they learned a weak encryption scheme that was easily breakable and of course if could not break because if it's just a narrow network is not Shamir or anything so I mean I look at that at this and I thought it's just unfair because if it's just a network it just has access to the ciphertext and could never break this this system so I tried to do something give a little edge to Eve so I used the chosen play text attack and Eve now choose two plain text and sent to Alice Alice and choose one and creeps and Bob etc so in this setup I ran and now with the neural network it could find though the one time pads this is the results is almost 100% of the time and yeah it was fun useless but fun and this was a kind of algorithm is a one-time pad with the with a permutation of the keys just that thank you