Burst uses a new algorithm for proof of hdd capacity (POC) mining. Miners pre-generate chunks of data known as 'plots' which are then saved to disk. The number of plots you store is effectively your mining speed. Every block the miner will skim through the saved plots, and come up with an amount of time until it is able to mine a block if another block hasn't yet been found. After reading through the plots is complete, your hardware can idle until the newest block.Miners generate and cache chunks of data known as 'plots', which are divided into 4096 portions known as 'scoops. Plots are generated by taking a public address and a nonce, then hashing it, pre-appending the resulting hash, repeating the hash-pre-append cycle many times, and then hashing the whole thing and xor'ing the last hash with the whole thing. Plots are staggered together so chunks of the same scoop number are together, then written to disk. Each block has a generation signature which is derived only from the previous block's generation signature and miner, so it is difficult to manipulate. When mining, the scoop number to be used for a block is derived from the generation signature and the block height, so the miner reads all relevant scoops(each plot will have 1 relevant scoop, and staggering allows for larger sequential read with less seeking) Only 0.024% of the stored data will need to be read each block. The generation signature is hashed with each scoop. 8 bytes are taken from the hash, then divided by a scaling factor (inverse difficulty). The resulting number is a number of seconds. If that many seconds passes since the last block without a new one, the address/nonce combination used to generate that plot/scoop is eligible to announce a new block. The miner's hardware can just sit idle until either that time or a new block.The address/nonce is included in the block as proof of eligibility, and the block is signed by that address. Technically, this mining process can be mined POW-style, however mining it as intended will yield thousands of times the hashrate, and your hardware will sit idle most of the time. Continuously hashing until a block is found is unnecessary, as waiting long enough will cause any nonce to eventually become valid.