Actually, it seems Hash-on-blockchain approach more forward thinking than vanilla cryptonight proof-of-work.
For now ASIC is problem for neither. When it becomes economically viable, the research and development would begin.
example:
Quote
True, we goes with different approach, but to the same goal.
The goal is not to make hash which is perfect for CPU.
The goal is to make "egalitarian proof of work", in other words: to provide almost equal voting rights to all participants.
Now traditionally approach is to have heavy hash-function, which is hard to implement in custom hardware. Cryptonote (CryptoNight) PoW hash(so called "slow_hash") goes on this way. It use some modern instructions sets coupled with memory consuming algorithm. Nowdays 2MB for ASIC seems to be almost impossible, but the world changing so fast now. And this finally could come to the same situation as Litecoin have now.
Our approach is to keep using modern instructions sets, but instead of making big scratchpad and then process whole pad for each hash call, we gonna use blockchain as one big source of random data, and will use pseudo-random parts of it for each hash.Two-phase hashing will protect from pre-calculation of all possible hashes(second phase hash randomly accessed data with salt). Blockchain after year of network's life will be consisited of about 1.5GB random data. In mining process all this data should be randomly accessible - and back to talking about ASICs - even if ASIC calculator for this hash will apear it will be almost impossible to grant paralel access of huge count of hardware calculators to to big scratchpad- due to DRAM specific.
The goal is not to make hash which is perfect for CPU.
The goal is to make "egalitarian proof of work", in other words: to provide almost equal voting rights to all participants.
Now traditionally approach is to have heavy hash-function, which is hard to implement in custom hardware. Cryptonote (CryptoNight) PoW hash(so called "slow_hash") goes on this way. It use some modern instructions sets coupled with memory consuming algorithm. Nowdays 2MB for ASIC seems to be almost impossible, but the world changing so fast now. And this finally could come to the same situation as Litecoin have now.
Our approach is to keep using modern instructions sets, but instead of making big scratchpad and then process whole pad for each hash call, we gonna use blockchain as one big source of random data, and will use pseudo-random parts of it for each hash.Two-phase hashing will protect from pre-calculation of all possible hashes(second phase hash randomly accessed data with salt). Blockchain after year of network's life will be consisited of about 1.5GB random data. In mining process all this data should be randomly accessible - and back to talking about ASICs - even if ASIC calculator for this hash will apear it will be almost impossible to grant paralel access of huge count of hardware calculators to to big scratchpad- due to DRAM specific.
The GPU is a small problem for BoolBerry. Scratchpad increase will reduce the gap. Hardfork is possible.
mining with MIC-like architecture on Monero is also more profitable than CPU
![Grin](https://bitcointalk.org/Smileys/default/grin.gif)