What is new is that it forces optimized miners to be good a neural net inference. (some reasons to do so are explained here: https://github.com/nf-dj/robocoin)
To do so the hashing function is replaced by using rounds of ternary matmuls, each round being NP hard to reverse (https://github.com/nf-dj/robocoin/blob/main/tens_pow.md).
In the future imagine robots with powerful NPUs mining during their sleep haha. (would have no chance mining sha256d or memory hard pows, with this pow optimizing miner == optimizing nn inference)